Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [Videocardz]AMD’s “Project F” is 232mm2 discrete GPU made in 14LPP process
New Posts  All Forums:Forum Nav:

[Videocardz]AMD’s “Project F” is 232mm2 discrete GPU made in 14LPP process - Page 6

post #51 of 99
Quote:
Originally Posted by sinholueiro View Post


The companies doesn't care about what you and me want in performance. They will just give you the same as always. get the 780Ti performance and calculate the gains from going to the 980. There you have the gains from going from GM200 to GP104. Barely changed arch from Maxwell and less nm to be able to put the same core count as a Titan X in the 980 die size and being able to get a big number of mid-sized chips from the waffle. It's just as simple as that.

The illusion of a GP104 being a top card do not work so well if it is not 20 to 40% faster that earlier top card. They may change strategy of cause.

White Whisper II
(15 items)
 
CPUMotherboardGraphicsRAM
4790k 4ghz Stock or 4.6 2133 or -HT 4.7 2133 ASUS Z97-PRO non sold my 970 Kingston fury 1866 
Hard DriveCoolingOSMonitor
Samsung 850 evo  Noctura D14 Windows 7 Fujitsu P24W-6 IPS 1920x1200 
MonitorPowerCase
Benq GW2760hs 1080p VA Corsair RM850 Fractal define R4 
  hide details  
Reply
White Whisper II
(15 items)
 
CPUMotherboardGraphicsRAM
4790k 4ghz Stock or 4.6 2133 or -HT 4.7 2133 ASUS Z97-PRO non sold my 970 Kingston fury 1866 
Hard DriveCoolingOSMonitor
Samsung 850 evo  Noctura D14 Windows 7 Fujitsu P24W-6 IPS 1920x1200 
MonitorPowerCase
Benq GW2760hs 1080p VA Corsair RM850 Fractal define R4 
  hide details  
Reply
post #52 of 99
Quote:
Originally Posted by zealord View Post

It is probably the GPU that AMD was talking about "bringing VR required GPU performance for lower price".

That card is probably like 150-199$ and should be around 290X +-5% levels of performance and consume like 100W~. I have nothing official, but I feel like I am not to far off with my estimations.


Samsung 14nm LPP has >2x density as compared to TSMC 28nm (slightly more)


So a 232mm sq might means u can pack smth like 480mm sq worth of 28nm transistors on the small die.

Given its also on a newer architectural like polaris, I am expecting a performance lvl at least on par with R9 Nano
Project Frostbite
(15 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core I7-6800K OC 4.2GHz @ 1.28 Vcore ASUS X99 Sabertooth Gigabyte Aorus GTX 1080 Ti 11G Corsair Vengeance White LED 4x8GB DDR4-3200MHz 
Hard DriveCoolingOSMonitor
Samsung Evo 850 1TB Corsair H100i V2 AiO Window 10 64-Bits Home Premium Acer Predator XB271HU 
KeyboardPowerCaseMouse
Tesoro Spectrum Mechanical Keyboard (Blue Switc... Thermaltake Toughpower 1000W Gold Phantek Enthoo Pro Full Acrylic Black Razer Diamondback 3G 
AudioAudioAudio
Audio Engine A2+ Speaker Audio Engine D1 DAC Audio Engine S8 Subwoofer 
  hide details  
Reply
Project Frostbite
(15 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core I7-6800K OC 4.2GHz @ 1.28 Vcore ASUS X99 Sabertooth Gigabyte Aorus GTX 1080 Ti 11G Corsair Vengeance White LED 4x8GB DDR4-3200MHz 
Hard DriveCoolingOSMonitor
Samsung Evo 850 1TB Corsair H100i V2 AiO Window 10 64-Bits Home Premium Acer Predator XB271HU 
KeyboardPowerCaseMouse
Tesoro Spectrum Mechanical Keyboard (Blue Switc... Thermaltake Toughpower 1000W Gold Phantek Enthoo Pro Full Acrylic Black Razer Diamondback 3G 
AudioAudioAudio
Audio Engine A2+ Speaker Audio Engine D1 DAC Audio Engine S8 Subwoofer 
  hide details  
Reply
post #53 of 99
Quote:
Originally Posted by G woodlogger View Post

The illusion of a GP104 being a top card do not work so well if it is not 20 to 40% faster that earlier top card. They may change strategy of cause.

Exactly. Good post.

Nvidia loves their loyal customers and they love screwing with them even more. Milking them with the illusion of always in need to have the best card.

If Jen-Hsun Huang could he'd release a new Nvidia card each month with a 5% performance increase.

Nvidia wants people to buy the GTX 1080, even if they already have a 980 Ti. If the 1080 doesn't turn out to be as good as people hoped then they can still resort to their old strategy of gimping old cards to make Pascal cards look better with new drivers like they did with Kepler cards after Maxwell came out biggrin.gif

Quote:
Originally Posted by guttheslayer View Post

Samsung 14nm LPP has >2x density as compared to TSMC 28nm (slightly more)


So a 232mm sq might means u can pack smth like 480mm sq worth of 28nm transistors on the small die.

Given its also on a newer architectural like polaris, I am expecting a performance lvl equivalent.to R9 Nano

Thanks for the post. R9 Nano is quite beefy. How much better than a 290X is it? 15-20%?
Although I feel like that is maybe too much for the new card. It would be great though and make me happy.
current rig
(4 items)
 
  
CPUGraphicsGraphicsKeyboard
2500k R9 290X Tri-X GTX 680 Filco Majestouch 2 Brown Ninja 
  hide details  
Reply
current rig
(4 items)
 
  
CPUGraphicsGraphicsKeyboard
2500k R9 290X Tri-X GTX 680 Filco Majestouch 2 Brown Ninja 
  hide details  
Reply
post #54 of 99
Quote:
Originally Posted by zealord View Post

Exactly. Good post.

Nvidia loves their loyal customers and they love screwing with them even more. Milking them with the illusion of always in need to have the best card.

If Jen-Hsun Huang could he'd release a new Nvidia card each month with a 5% performance increase.

Nvidia wants people to buy the GTX 1080, even if they already have a 980 Ti. If the 1080 doesn't turn out to be as good as people hoped then they can still resort to their old strategy of gimping old cards to make Pascal cards look better with new drivers like they did with Kepler cards after Maxwell came out biggrin.gif
Thanks for the post. R9 Nano is quite beefy. How much better than a 290X is it? 15-20%?
Although I feel like that is maybe too much for the new card. It would be great though and make me happy.


R9 Nano is beefy but its severely underclocked, so cant really compared in term of size alone, clock speed need to factor in.


But given the leap in both architectural and process improvement, its not hard to assume the perf-per-watt is at least 2x better. So under the senario, a new GPU that has same performance as R9 Nano effectively only has a TDP of 90W (direct scaling). A 232mm GPU is definitely capable of >90W of processing power, which means the new GPU can easily perform as great as a R9 Nano.
Project Frostbite
(15 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core I7-6800K OC 4.2GHz @ 1.28 Vcore ASUS X99 Sabertooth Gigabyte Aorus GTX 1080 Ti 11G Corsair Vengeance White LED 4x8GB DDR4-3200MHz 
Hard DriveCoolingOSMonitor
Samsung Evo 850 1TB Corsair H100i V2 AiO Window 10 64-Bits Home Premium Acer Predator XB271HU 
KeyboardPowerCaseMouse
Tesoro Spectrum Mechanical Keyboard (Blue Switc... Thermaltake Toughpower 1000W Gold Phantek Enthoo Pro Full Acrylic Black Razer Diamondback 3G 
AudioAudioAudio
Audio Engine A2+ Speaker Audio Engine D1 DAC Audio Engine S8 Subwoofer 
  hide details  
Reply
Project Frostbite
(15 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core I7-6800K OC 4.2GHz @ 1.28 Vcore ASUS X99 Sabertooth Gigabyte Aorus GTX 1080 Ti 11G Corsair Vengeance White LED 4x8GB DDR4-3200MHz 
Hard DriveCoolingOSMonitor
Samsung Evo 850 1TB Corsair H100i V2 AiO Window 10 64-Bits Home Premium Acer Predator XB271HU 
KeyboardPowerCaseMouse
Tesoro Spectrum Mechanical Keyboard (Blue Switc... Thermaltake Toughpower 1000W Gold Phantek Enthoo Pro Full Acrylic Black Razer Diamondback 3G 
AudioAudioAudio
Audio Engine A2+ Speaker Audio Engine D1 DAC Audio Engine S8 Subwoofer 
  hide details  
Reply
post #55 of 99
Quote:
Originally Posted by zealord View Post

Well I disagree.
What do you define "as always"? Is "as always" just your recent memory? The jump from GTX 580 -> 680 was 35% performance increase and they went from a 520mm² to 294mm² die. What makes you think that going from Maxwell to Pascal is like going from Kepler to Maxwell instead of Fermi to Kepler? Fermi to Kepler makes much more sense.
What do you base that on? 780 Ti and 980 are both 28nm cards. GM200 to GP104 is a jump from 28nm to 16/14nm.

You know that the 780 Ti to 980 is only a very small jump? ( http://www.anandtech.com/show/8526/nvidia-geforce-gtx-980-review/14 ) It is like 2-3% better. You honestly expect a jump that small? May I ask you an honest question? Are you trying to downplay the GP104 so you would feel better if you buy it when it releases? Are we talking about different cards?
I am talking about the card that I expect to be the GTX "1080".
Why would you expect the GTX 1080 to be only 2-3% better than the 980 Ti?

(my post might sound offensive, but I am actually curious)

I'm based in the point that we won't change the arch like going from 580 to 680, from Fermi to Kepler. Pascal is just Maxwell at 16nm. Can be some tweaks but it's just the same thing. Ang going to less nm only helps to pack more cores and contain the power consumption. There's now a time where going to less nm doesn't improve performance per se. And I was talking about the differences now, being the 780Ti more or less like a stock 970. The 980 has rise since launch. I am not trying to downplay the GP104, I'm trying to be realistic. I don't have any plans in buying the GP104 if that makes you feel better. And yes, I expect the cut down GP104 to be that better over the 980Ti. The full is maybe 10% better than cut-down and 980Ti.
X99
(12 items)
 
  
CPUMotherboardGraphicsRAM
5820k Asus X99A GTX 1080 GSkill Ripjaws V 
Hard DriveCoolingOSMonitor
850 EVO 250GB Corsair H110i Arch Linux Dell P2416D 
KeyboardPowerCaseMouse
Rantopad MXX RGB Corsair RM750x Corsair Obsidian 750D Razer DeathAdder 
  hide details  
Reply
X99
(12 items)
 
  
CPUMotherboardGraphicsRAM
5820k Asus X99A GTX 1080 GSkill Ripjaws V 
Hard DriveCoolingOSMonitor
850 EVO 250GB Corsair H110i Arch Linux Dell P2416D 
KeyboardPowerCaseMouse
Rantopad MXX RGB Corsair RM750x Corsair Obsidian 750D Razer DeathAdder 
  hide details  
Reply
post #56 of 99
Quote:
Originally Posted by sinholueiro View Post

I'm based in the point that we won't change the arch like going from 580 to 680, from Fermi to Kepler. Pascal is just Maxwell at 16nm. Can be some tweaks but it's just the same thing. Ang going to less nm only helps to pack more cores and contain the power consumption. There's now a time where going to less nm doesn't improve performance per se. And I was talking about the differences now, being the 780Ti more or less like a stock 970. The 980 has rise since launch. I am not trying to downplay the GP104, I'm trying to be realistic. I don't have any plans in buying the GP104 if that makes you feel better. And yes, I expect the cut down GP104 to be that better over the 980Ti. The full is maybe 10% better than cut-down and 980Ti.


It was mentioned that pascal give 2x perf per watt improvement based on architecture change alone. I doubt it was just maxwell architecture. I think the new architecture alone will boost at least 20% more perf per-core per-clock
Edited by guttheslayer - 2/14/16 at 8:20am
Project Frostbite
(15 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core I7-6800K OC 4.2GHz @ 1.28 Vcore ASUS X99 Sabertooth Gigabyte Aorus GTX 1080 Ti 11G Corsair Vengeance White LED 4x8GB DDR4-3200MHz 
Hard DriveCoolingOSMonitor
Samsung Evo 850 1TB Corsair H100i V2 AiO Window 10 64-Bits Home Premium Acer Predator XB271HU 
KeyboardPowerCaseMouse
Tesoro Spectrum Mechanical Keyboard (Blue Switc... Thermaltake Toughpower 1000W Gold Phantek Enthoo Pro Full Acrylic Black Razer Diamondback 3G 
AudioAudioAudio
Audio Engine A2+ Speaker Audio Engine D1 DAC Audio Engine S8 Subwoofer 
  hide details  
Reply
Project Frostbite
(15 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core I7-6800K OC 4.2GHz @ 1.28 Vcore ASUS X99 Sabertooth Gigabyte Aorus GTX 1080 Ti 11G Corsair Vengeance White LED 4x8GB DDR4-3200MHz 
Hard DriveCoolingOSMonitor
Samsung Evo 850 1TB Corsair H100i V2 AiO Window 10 64-Bits Home Premium Acer Predator XB271HU 
KeyboardPowerCaseMouse
Tesoro Spectrum Mechanical Keyboard (Blue Switc... Thermaltake Toughpower 1000W Gold Phantek Enthoo Pro Full Acrylic Black Razer Diamondback 3G 
AudioAudioAudio
Audio Engine A2+ Speaker Audio Engine D1 DAC Audio Engine S8 Subwoofer 
  hide details  
Reply
post #57 of 99
Quote:
Originally Posted by sinholueiro View Post

I'm based in the point that we won't change the arch like going from 580 to 680, from Fermi to Kepler. Pascal is just Maxwell at 16nm. Can be some tweaks but it's just the same thing. Ang going to less nm only helps to pack more cores and contain the power consumption. There's now a time where going to less nm doesn't improve performance per se. And I was talking about the differences now, being the 780Ti more or less like a stock 970. The 980 has rise since launch. I am not trying to downplay the GP104, I'm trying to be realistic. I don't have any plans in buying the GP104 if that makes you feel better. And yes, I expect the cut down GP104 to be that better over the 980Ti. The full is maybe 10% better than cut-down and 980Ti.

All right. Thanks for your honesty

I thought that from an architecture standpoint that Pascal was better than Maxwell by like 2x perf/watt. Do you have any good articles or something for me that I can read about going from Maxwell to Pascal?
current rig
(4 items)
 
  
CPUGraphicsGraphicsKeyboard
2500k R9 290X Tri-X GTX 680 Filco Majestouch 2 Brown Ninja 
  hide details  
Reply
current rig
(4 items)
 
  
CPUGraphicsGraphicsKeyboard
2500k R9 290X Tri-X GTX 680 Filco Majestouch 2 Brown Ninja 
  hide details  
Reply
post #58 of 99
Quote:
Originally Posted by zealord View Post

Nvidia wants people to buy the GTX 1080, even if they already have a 980 Ti. If the 1080 doesn't turn out to be as good as people hoped then they can still resort to their old strategy of gimping old cards to make Pascal cards look better with new drivers like they did with Kepler cards after Maxwell came out biggrin.gif
Imagine how beefy that new tessellation pipeline would be though to only see a 20-30% performance hit from128x tess. tongue.gif

On a serious note, with a new architecture who to say this thing wont pull ahead of gcn's low clocking history. Say equal to a 290 at 900mhz but it boost to 1300 with overclocking to 1400mhz. Or is this just a smaller refresh tahiti to hawaii style with "more cores" being able to be stuffed in a smaller area?
thy gaming wench
(15 items)
 
server(old gamer)
(10 items)
 
 
CPUMotherboardGraphicsRAM
AMD FX-8350 Vishera 4.0GHz (4.2GHz Turbo) Socke...  ASUS SABERTOOTH 990FX R2.0 AM3+ AMD 990FX SATA...  SAPPHIRE 100354OCL Radeon HD 7870 GHz Edition ... Mushkin Enhanced Redline 16GB (4 x 4GB) 240-Pin... 
Hard DriveOptical DriveCoolingOS
samsung 840 256GB 500GB seagate some cheap asus  Thermaltake CLW0217 Water 2.0 Extreme/All-In-O... winduhs pro 64 
MonitorKeyboardPowerCase
ASUS VS Series VS247H-P Black 23.6" 2ms LED Ba... slick looking logitech for 50ish  RAIDMAX RX-1000AE 1000W ATX12V v2.3 / EPS12V S...  IN WIN Dragon Rider Black 1.0 - 0.8mm SECC Ste... 
MouseMouse PadAudio
some 25 dollar logitech  pfft Logitech 5.1 surround + sub  
CPUMotherboardGraphicsRAM
intel core 2 duo e8400 biostar tpower i45 some old 4000 series ati card that i didnt want... 8gb of gskill ddr2 800 
Hard DriveCoolingOSMonitor
3x1TB in raid z1 and 4x3TB in raidz2 cooler master hyper 212 freenas 8.3.1(freeBSD) headless 
PowerCase
corair 750 watt cooler master centurion 5 
  hide details  
Reply
thy gaming wench
(15 items)
 
server(old gamer)
(10 items)
 
 
CPUMotherboardGraphicsRAM
AMD FX-8350 Vishera 4.0GHz (4.2GHz Turbo) Socke...  ASUS SABERTOOTH 990FX R2.0 AM3+ AMD 990FX SATA...  SAPPHIRE 100354OCL Radeon HD 7870 GHz Edition ... Mushkin Enhanced Redline 16GB (4 x 4GB) 240-Pin... 
Hard DriveOptical DriveCoolingOS
samsung 840 256GB 500GB seagate some cheap asus  Thermaltake CLW0217 Water 2.0 Extreme/All-In-O... winduhs pro 64 
MonitorKeyboardPowerCase
ASUS VS Series VS247H-P Black 23.6" 2ms LED Ba... slick looking logitech for 50ish  RAIDMAX RX-1000AE 1000W ATX12V v2.3 / EPS12V S...  IN WIN Dragon Rider Black 1.0 - 0.8mm SECC Ste... 
MouseMouse PadAudio
some 25 dollar logitech  pfft Logitech 5.1 surround + sub  
CPUMotherboardGraphicsRAM
intel core 2 duo e8400 biostar tpower i45 some old 4000 series ati card that i didnt want... 8gb of gskill ddr2 800 
Hard DriveCoolingOSMonitor
3x1TB in raid z1 and 4x3TB in raidz2 cooler master hyper 212 freenas 8.3.1(freeBSD) headless 
PowerCase
corair 750 watt cooler master centurion 5 
  hide details  
Reply
post #59 of 99
-delete-
Project Frostbite
(15 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core I7-6800K OC 4.2GHz @ 1.28 Vcore ASUS X99 Sabertooth Gigabyte Aorus GTX 1080 Ti 11G Corsair Vengeance White LED 4x8GB DDR4-3200MHz 
Hard DriveCoolingOSMonitor
Samsung Evo 850 1TB Corsair H100i V2 AiO Window 10 64-Bits Home Premium Acer Predator XB271HU 
KeyboardPowerCaseMouse
Tesoro Spectrum Mechanical Keyboard (Blue Switc... Thermaltake Toughpower 1000W Gold Phantek Enthoo Pro Full Acrylic Black Razer Diamondback 3G 
AudioAudioAudio
Audio Engine A2+ Speaker Audio Engine D1 DAC Audio Engine S8 Subwoofer 
  hide details  
Reply
Project Frostbite
(15 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core I7-6800K OC 4.2GHz @ 1.28 Vcore ASUS X99 Sabertooth Gigabyte Aorus GTX 1080 Ti 11G Corsair Vengeance White LED 4x8GB DDR4-3200MHz 
Hard DriveCoolingOSMonitor
Samsung Evo 850 1TB Corsair H100i V2 AiO Window 10 64-Bits Home Premium Acer Predator XB271HU 
KeyboardPowerCaseMouse
Tesoro Spectrum Mechanical Keyboard (Blue Switc... Thermaltake Toughpower 1000W Gold Phantek Enthoo Pro Full Acrylic Black Razer Diamondback 3G 
AudioAudioAudio
Audio Engine A2+ Speaker Audio Engine D1 DAC Audio Engine S8 Subwoofer 
  hide details  
Reply
post #60 of 99
Quote:
Originally Posted by guttheslayer View Post

Nope. It was mentioned that pascal give 2x perf per watt improvement based on architecture change alone. It was independent of process node change
Where exactly was this? The only thing I've seen in regards to that was some research papers stating roughly 2x FP32/W.
Quote:
Originally Posted by Darklyric View Post

Imagine how beefy that new tessellation pipeline would be though to only see a 20-30% performance hit from128x tess. tongue.gif

On a serious note, with a new architecture who to say this thing wont pull ahead of gcn's low clocking history. Say equal to a 290 at 900mhz but it boost to 1300 with overclocking to 1400mhz. Or is this just a smaller refresh tahiti to hawaii style with "more cores" being able to be stuffed in a smaller area?
Polaris seems a bit larger revision. I'd expect clock speed to go up solely due to the node change. Pretty sure at least the geometry processors and L2 had changes to them.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Rumors and Unconfirmed Articles
Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [Videocardz]AMD’s “Project F” is 232mm2 discrete GPU made in 14LPP process