Overclock.net › Forums › Industry News › Hardware News › [wccftech] GeForce 800 Series Arrives in Second Half of 2014
New Posts  All Forums:Forum Nav:

[wccftech] GeForce 800 Series Arrives in Second Half of 2014 - Page 93

post #921 of 1139
Looking at Maxwell's performance per watt with the two released cards, if that ratio increase is consistent in the midrange and high-end models then Nvidia doesn't need 20nm right now, especially since yields are probably low and transitioning for what I've read are minimal gains is probably not cost effective. They might as well wait another year for 16nm FinFet which has the benefits of a full node transition in order to release 2nd generation Maxwell.

If a future GTX 860 / 860 Ti can reach GTX 680 and surpass GTX 770 levels of performance on a 120w TDP, 200w+ - 250w cards at the high-end can easily surpass the GTX 780 Ti.

I think with the Nvidia claims that a 128 CUDA core SMM (Maxwell) has 90% of the performance of a 192 CUDA core SMX (Kepler), and they seem to be true based on the current cards, they have basically ensured they can launch the whole first generation Maxwell on 28nm.


Here are my predictions:


Mid-range: 4 MB of L2 cache, 192-bit memory bus, 3 GB of 7 Ghz GDDR5 or 256-bit memory bus, 4 GB of 6 Ghz GDDR5, the large amount of L2 cache in Maxwell compared to Kepler lowers the need for as much memory bandwidth, my doubt is how much VRAM they want to give the midrange cards.

GTX 860: 1024 Cuda cores, faster than the GTX 760 and GTX 670, depending on clocks might be as fast as GTX 680; price: $ 230 - $250

GTX 860 Ti: 1280 Cuda cores, faster than a GTX 770; price: $ 299;



High-end: 8 MB of L2 cache, 320-bit memory bus, 5 GB of 7 Ghz GDDR5

GTX 870: 2048 Cuda cores, probably 4 -5 % slower than a GTX 780 Ti or about as fast, depending on stock clockspeed; price: $399

Edit: GTX 880: 2432 Cuda cores, 16% faster than the GTX 780 Ti; price: $499

More probably: GTX 880: 2560 Cuda cores, 20% faster than the GTX 780 Ti; price $499


Released later in order to extend the 1st generation's life:

GTX 880 Ti: 2816 Cuda cores, 35% faster than the GTX 780 Ti; price: $649


Having these projections in mind, a June release at the earliest, it makes sense to release at the same time as Haswell-e in order to benefit from the upgrade waive.
Edited by tpi2007 - 3/15/14 at 10:27pm
 
Metro 2033 review
Metro 2033
CPUMotherboardGraphicsRAM
Core i7-3820 Asus Sabertooth X79 MSI GTX 750 Ti TF Gaming 16 GB Corsair DDR3 1866 Mhz Dominator 
Hard DriveOptical DriveCoolingOS
Samsung SSD 830 128GB + WD Caviar Black 1TB Sony Optiarc DVD-RW Corsair A70 + Noiseblocker M12-P Windows 7 Home Premium 64-bit 
MonitorKeyboardPowerCase
BenQ RL2455HM Cooler Master Octane Corsair AX750 Professional Modular 80 Plus Gold Cooler Master HAF 912 Plus 
Mouse
Cooler Master Octane 
  hide details  
Reply
 
Metro 2033 review
Metro 2033
CPUMotherboardGraphicsRAM
Core i7-3820 Asus Sabertooth X79 MSI GTX 750 Ti TF Gaming 16 GB Corsair DDR3 1866 Mhz Dominator 
Hard DriveOptical DriveCoolingOS
Samsung SSD 830 128GB + WD Caviar Black 1TB Sony Optiarc DVD-RW Corsair A70 + Noiseblocker M12-P Windows 7 Home Premium 64-bit 
MonitorKeyboardPowerCase
BenQ RL2455HM Cooler Master Octane Corsair AX750 Professional Modular 80 Plus Gold Cooler Master HAF 912 Plus 
Mouse
Cooler Master Octane 
  hide details  
Reply
post #922 of 1139
Yes, yes they do.... I am quite sure. New games spur gpu sales just as they do for consoles.
The Sig Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600K @ 4.xghz MSI Z68A-GD65 (G3) MSI GTX 970 Gaming SLI @ 1504mhz core/8020mhz m... 16GB DDR3 2x 8gb @ 1866mhz C8 
Hard DriveHard DriveOptical DriveCooling
240GB Sandisk Extreme A few 2tb hdd's. DVDRW Corsair H75 CLC on the CPU 
OSMonitorKeyboardPower
Windows 8.1 Pro 64-bit ACER B326hk 4k 60hz 3840x2160 IPS sst display Coolermaster Cherry MX Brown switch mechanical ... Corsair TX 950w 
CaseMouseMouse PadAudio
Fractal Design Define R3 Black Pearl Logitech MX518 EverGlide Titan "Medium" Soundblaster Z PCIE 
  hide details  
Reply
The Sig Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600K @ 4.xghz MSI Z68A-GD65 (G3) MSI GTX 970 Gaming SLI @ 1504mhz core/8020mhz m... 16GB DDR3 2x 8gb @ 1866mhz C8 
Hard DriveHard DriveOptical DriveCooling
240GB Sandisk Extreme A few 2tb hdd's. DVDRW Corsair H75 CLC on the CPU 
OSMonitorKeyboardPower
Windows 8.1 Pro 64-bit ACER B326hk 4k 60hz 3840x2160 IPS sst display Coolermaster Cherry MX Brown switch mechanical ... Corsair TX 950w 
CaseMouseMouse PadAudio
Fractal Design Define R3 Black Pearl Logitech MX518 EverGlide Titan "Medium" Soundblaster Z PCIE 
  hide details  
Reply
post #923 of 1139
Here is my theory from mid February.... posted Feb 15th. http://hardforum.com/showthread.php?t=1806694
Quote:
Originally Posted by GoldenTiger;1040629439 
I'm gonna go off the deep end here with a detailed guesswork at what we might see with GM200 at 20nm. .

8v51.jpg

This slide says the case for Maxwell's first generation is reportedly, for the same performance level we should be seeing half the wattage used even on 28nm, and that per-core we are going to see a 35% gain (by their claims, at this point of course). So a card that would take Kepler 300w can be done in a 150w or less envelope is their claim, on 28nm even. That leaves a huge amount of extra power to continue stuffing in more cores and increasing clocks. Now add in that the Big Maxwell is going to be (GM200/204) on 20nm, and that improves power efficiency and die size usable even further. Now add in that this is Maxwell FIRST generation, for the GM107, and that GM200/204 are SECOND generation by all known info.... and we have a recipe for easily doubling the performance at least, if their claims are anywhere near true.

So, time for some math based on the theoretical and what is a real product launching in a couple of days (GTX 750 Ti).

GM107 has 5 SMM units containing 128 cuda cores, in each GPC. One GPC is what GM107 is using. It uses 60 watts for performance that is only beaten out by a GTX 660 by about 12% per the leaked benches. A GTX 660 has 960 Kepler cores, while a GM107 has 640 Maxwell cores. A GTX 660 uses a TDP rating of 140 watts. See an interesting number here?

960 is 50% more cores than the 640 ones Maxwell is using for the 750Ti. Now, it is around 12% slower there.... see the magic number close by? Nvidia claims a 35% performance increase PER CORE and additionally that the cores there will use only around half the power overall in TDP. Now extrapolate on some napkin math: the GM107 only has a 128-bit bus. That means the bandwidth efficiency is greatly improved because a GTX 650 Ti Boost is a huge amount above a GTX 650 Ti (192-bit vs. 128-bit bus widths and basically the same otherwise).

So it's safe to say that with 50% more bandwidth and 50% more cores, a GTX 660 only performing ~12% higher than a Maxwell part with slightly higher clocks is incredibly impressive. Now let's scale! We know at 28nm lithography that GM107 is around 148 square millimeters for the die size. Let's COMPLETELY IGNORE 20NM for a second here on the power savings and size! Forget about it for a minute. It would be extremely easy to see, since even on 28nm the reticule size is around 570mm2, that they could use 3 GPC units on 28nm taking around 420-430mm2 with this imaginary chip that would have 15 SMM units. 15 SMM units times 128 per unit would mean 1920 Maxwell cores.

Pretend their scheduler is great and the performance scales well with core count and clocks, and that they kept the idea of triple everything there in this hypothetical, non-existent card that is an illustration only. So we'd have a 384 bit bus with 1920 Maxwell cores, probably 7ghz memory speed of GDDR5 like Kepler does at least, and a TDP that fits inside of 200 watts. Now let's say that you only get about 75% scaling from core count here, which is reasonable even though Kepler scales pretty linearly, but it's a new architecture with Maxwell, so let's make the safe assumption. So a GTX 660 performs 12% better than a GM107 with 640 cores. Triple the core count there with our rough napkin math again with everything else and you would have a card performing around the same as GK110 fully unlocked by that theory-crafting, at least, and it has better potential for higher clocks thanks to the lower power usage.

However, in reality, we know they are going 20nm. This allows for even more power savings. This also allows for many more transistors per square mm on the die. So pretend they want to go for a 520mm2 chip on 20nm, keep costs down a tad for Big Maxwell and improve yields per wafer. According to released documents such as this: http://www.cadence.com/rl/Resources/overview/20nm_qa.pdf we can expect to see transistor counts possible of 8-12 billion. GK110 is 7.1 billion transistors. Using that as a point of reference, let's scale 28nm GK110 to 28nm Maxwell GM107: we need less memory controllers and pad space, so we can safely come up with a number in the neighborhood, considering the 148mm2 die size compared to GK110's 551mm2 size. At 20nm, you will be able to fit upwards of 11-12 billion transistors for a high-end part. For 28nm the 148mm2 die size indicates roughly 1.7-1.8 billion transistors with the 128-bit bus.

So now we have a decent number here: we know that Maxwell at 28nm in GM107 form is taking about 2 billion transistors to perform at a level about 89% as fast as a GK106. Let's use this as a base for the next part of this thought exercise smile.gif.

So 2 billion.... it's safe to say they could fit 7 GPC's at 20nm easily since 20nm should provide roughly a 2x density shrink in die size used per transistor, very easily, and we wouldn't need to duplicate memory controllers beyond 3x of the 28nm design's if we went for a 384-bit bus. At 20nm let's say they went for a 384-bit bus, to feed 7 GPC's worth of cores since the architecture is more bandwidth-efficient than Kepler, clearly by far. That'd give us 4480 Maxwell cores which already are much more bandwidth-efficient, so it probably would be plenty well-fed on that end of things.

Power-wise we'd be looking at an envelope of, 250 watts since Maxwell GM107 at 28nm takes a full-card power of 60w for 640 cores. That means probably around 45w for the GPU itself, allowing 10w for the GDDR5 and 5w for the fan and other circuitry. The move to 20nm will improve power efficiency drastically, due to the shorter gate lengths. So it's fair to say they could fit seven of those inside of that envelope, more than easily.

Again, let's go for a linear scaling factor of approximately 75% for the core count improvements... so we have a GTX 660 GK106 card we will use and compare to a GK110 780 Ti. A GTX 660 at 1080p is able to pull about 50% of the performance of a fully-unlocked, 780 Ti (source: http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_780_Ti/27.html).

Now we know each SMM will provide around 90% of that performance, and revision 2.0 (second gen) on 20nm will probably be closer to a 100% figure. So let's conservatively say each SMM-based GPC results in a real-world performance of a GTX 660. In other words, around half of a 780 Ti. Now let's conservatively also say we only see the benefit of 75% of the cores when scaling it to 7 GPC's or 35 SMM's. 35 SMM's would be, as you recall 4480 maxwell cores. Multiply the performance of the 50% card by 7 and we'd have 350% (or 2.5x faster than) of the performance of a full GK110. However, let's now apply the 75% rough rule and we come up with a much more reasonable 262.5% of the performance, or 2.6x as fast as (1.6x faster than) a GTX 780 Ti.

My predictions, therefore, are that we will see a Big Maxwell on 20nm with at minimum twice the performance of a GTX 780 Ti, and by current rumors it is due this year. Add in that Maxwell will be able to clock higher at 20nm (I based those power figures off of the numbers above which were of a card at 28nm with a 1085mhz GPU clock, and realistically they can probably fit a stock core speed of 1100-1150mhz of this big a chip in then. Thus my predictions are we would see a 450-460mm2 die size for this hypothetical card, at 20nm, with a 250-260w TDP rating and thanks to the 2x transistor density, approximately 9.2b transistors.

Napkin math, for sure, but worth thinking about, eh?
The Sig Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600K @ 4.xghz MSI Z68A-GD65 (G3) MSI GTX 970 Gaming SLI @ 1504mhz core/8020mhz m... 16GB DDR3 2x 8gb @ 1866mhz C8 
Hard DriveHard DriveOptical DriveCooling
240GB Sandisk Extreme A few 2tb hdd's. DVDRW Corsair H75 CLC on the CPU 
OSMonitorKeyboardPower
Windows 8.1 Pro 64-bit ACER B326hk 4k 60hz 3840x2160 IPS sst display Coolermaster Cherry MX Brown switch mechanical ... Corsair TX 950w 
CaseMouseMouse PadAudio
Fractal Design Define R3 Black Pearl Logitech MX518 EverGlide Titan "Medium" Soundblaster Z PCIE 
  hide details  
Reply
The Sig Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600K @ 4.xghz MSI Z68A-GD65 (G3) MSI GTX 970 Gaming SLI @ 1504mhz core/8020mhz m... 16GB DDR3 2x 8gb @ 1866mhz C8 
Hard DriveHard DriveOptical DriveCooling
240GB Sandisk Extreme A few 2tb hdd's. DVDRW Corsair H75 CLC on the CPU 
OSMonitorKeyboardPower
Windows 8.1 Pro 64-bit ACER B326hk 4k 60hz 3840x2160 IPS sst display Coolermaster Cherry MX Brown switch mechanical ... Corsair TX 950w 
CaseMouseMouse PadAudio
Fractal Design Define R3 Black Pearl Logitech MX518 EverGlide Titan "Medium" Soundblaster Z PCIE 
  hide details  
Reply
post #924 of 1139
So when do you all think Denver will rear its game changing, glorious, sexy, amazing, put everyone on their asses, head? 20nm, or 16nm?
IT'S WHITE
(21 items)
 
 
CPUMotherboardGraphicsGraphics
4930k Rampage iv extreme Black Edition GTX Titan  GTX Titan 
GraphicsRAMHard DriveHard Drive
GTX Titan GskillZ Trident X Samsung EVO samsung 840 evo 
Optical DriveCoolingCoolingOS
Asus External Blue Ray drive 4x 480x60mm Swiftech MCP35x2 Win 8.1.2 
MonitorMonitorMonitorPower
AOC AOC AOC EVGA G2 1300w 
CaseMouseMouse PadAudio
STH10 RAT7 what is this?  Vali / Modi :(  
Other
Corsair HX850 
  hide details  
Reply
IT'S WHITE
(21 items)
 
 
CPUMotherboardGraphicsGraphics
4930k Rampage iv extreme Black Edition GTX Titan  GTX Titan 
GraphicsRAMHard DriveHard Drive
GTX Titan GskillZ Trident X Samsung EVO samsung 840 evo 
Optical DriveCoolingCoolingOS
Asus External Blue Ray drive 4x 480x60mm Swiftech MCP35x2 Win 8.1.2 
MonitorMonitorMonitorPower
AOC AOC AOC EVGA G2 1300w 
CaseMouseMouse PadAudio
STH10 RAT7 what is this?  Vali / Modi :(  
Other
Corsair HX850 
  hide details  
Reply
post #925 of 1139
Quote:
Originally Posted by skupples View Post

So when do you all think Denver will rear its game changing, glorious, sexy, amazing, put everyone on their asses, head? 20nm, or 16nm?

20nm is going to be short lived by all accounts, so it's really a matter of timing. My hunch would be 16nm but that's all a pure guess.
The Sig Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600K @ 4.xghz MSI Z68A-GD65 (G3) MSI GTX 970 Gaming SLI @ 1504mhz core/8020mhz m... 16GB DDR3 2x 8gb @ 1866mhz C8 
Hard DriveHard DriveOptical DriveCooling
240GB Sandisk Extreme A few 2tb hdd's. DVDRW Corsair H75 CLC on the CPU 
OSMonitorKeyboardPower
Windows 8.1 Pro 64-bit ACER B326hk 4k 60hz 3840x2160 IPS sst display Coolermaster Cherry MX Brown switch mechanical ... Corsair TX 950w 
CaseMouseMouse PadAudio
Fractal Design Define R3 Black Pearl Logitech MX518 EverGlide Titan "Medium" Soundblaster Z PCIE 
  hide details  
Reply
The Sig Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600K @ 4.xghz MSI Z68A-GD65 (G3) MSI GTX 970 Gaming SLI @ 1504mhz core/8020mhz m... 16GB DDR3 2x 8gb @ 1866mhz C8 
Hard DriveHard DriveOptical DriveCooling
240GB Sandisk Extreme A few 2tb hdd's. DVDRW Corsair H75 CLC on the CPU 
OSMonitorKeyboardPower
Windows 8.1 Pro 64-bit ACER B326hk 4k 60hz 3840x2160 IPS sst display Coolermaster Cherry MX Brown switch mechanical ... Corsair TX 950w 
CaseMouseMouse PadAudio
Fractal Design Define R3 Black Pearl Logitech MX518 EverGlide Titan "Medium" Soundblaster Z PCIE 
  hide details  
Reply
post #926 of 1139
Quote:
Originally Posted by GoldenTiger View Post

20nm is going to be short lived by all accounts, so it's really a matter of timing. My hunch would be 16nm but that's all a pure guess.

That is what iv'e bean leaning towards as well. It would be funny if they end up just skipping 20nm all together... I thought 20nm also required the FinFet process, but that comes from the Hawaii speculation threads, and we all know how intelligent those folks are.
IT'S WHITE
(21 items)
 
 
CPUMotherboardGraphicsGraphics
4930k Rampage iv extreme Black Edition GTX Titan  GTX Titan 
GraphicsRAMHard DriveHard Drive
GTX Titan GskillZ Trident X Samsung EVO samsung 840 evo 
Optical DriveCoolingCoolingOS
Asus External Blue Ray drive 4x 480x60mm Swiftech MCP35x2 Win 8.1.2 
MonitorMonitorMonitorPower
AOC AOC AOC EVGA G2 1300w 
CaseMouseMouse PadAudio
STH10 RAT7 what is this?  Vali / Modi :(  
Other
Corsair HX850 
  hide details  
Reply
IT'S WHITE
(21 items)
 
 
CPUMotherboardGraphicsGraphics
4930k Rampage iv extreme Black Edition GTX Titan  GTX Titan 
GraphicsRAMHard DriveHard Drive
GTX Titan GskillZ Trident X Samsung EVO samsung 840 evo 
Optical DriveCoolingCoolingOS
Asus External Blue Ray drive 4x 480x60mm Swiftech MCP35x2 Win 8.1.2 
MonitorMonitorMonitorPower
AOC AOC AOC EVGA G2 1300w 
CaseMouseMouse PadAudio
STH10 RAT7 what is this?  Vali / Modi :(  
Other
Corsair HX850 
  hide details  
Reply
post #927 of 1139
Quote:
Originally Posted by GoldenTiger View Post

Here is my theory from mid February.... posted Feb 15th. http://hardforum.com/showthread.php?t=1806694
Warning: Huge napkin! (Click to show)
Quote:
Originally Posted by GoldenTiger;1040629439 
I'm gonna go off the deep end here with a detailed guesswork at what we might see with GM200 at 20nm. .

8v51.jpg

This slide says the case for Maxwell's first generation is reportedly, for the same performance level we should be seeing half the wattage used even on 28nm, and that per-core we are going to see a 35% gain (by their claims, at this point of course). So a card that would take Kepler 300w can be done in a 150w or less envelope is their claim, on 28nm even. That leaves a huge amount of extra power to continue stuffing in more cores and increasing clocks. Now add in that the Big Maxwell is going to be (GM200/204) on 20nm, and that improves power efficiency and die size usable even further. Now add in that this is Maxwell FIRST generation, for the GM107, and that GM200/204 are SECOND generation by all known info.... and we have a recipe for easily doubling the performance at least, if their claims are anywhere near true.

So, time for some math based on the theoretical and what is a real product launching in a couple of days (GTX 750 Ti).

GM107 has 5 SMM units containing 128 cuda cores, in each GPC. One GPC is what GM107 is using. It uses 60 watts for performance that is only beaten out by a GTX 660 by about 12% per the leaked benches. A GTX 660 has 960 Kepler cores, while a GM107 has 640 Maxwell cores. A GTX 660 uses a TDP rating of 140 watts. See an interesting number here?

960 is 50% more cores than the 640 ones Maxwell is using for the 750Ti. Now, it is around 12% slower there.... see the magic number close by? Nvidia claims a 35% performance increase PER CORE and additionally that the cores there will use only around half the power overall in TDP. Now extrapolate on some napkin math: the GM107 only has a 128-bit bus. That means the bandwidth efficiency is greatly improved because a GTX 650 Ti Boost is a huge amount above a GTX 650 Ti (192-bit vs. 128-bit bus widths and basically the same otherwise).

So it's safe to say that with 50% more bandwidth and 50% more cores, a GTX 660 only performing ~12% higher than a Maxwell part with slightly higher clocks is incredibly impressive. Now let's scale! We know at 28nm lithography that GM107 is around 148 square millimeters for the die size. Let's COMPLETELY IGNORE 20NM for a second here on the power savings and size! Forget about it for a minute. It would be extremely easy to see, since even on 28nm the reticule size is around 570mm2, that they could use 3 GPC units on 28nm taking around 420-430mm2 with this imaginary chip that would have 15 SMM units. 15 SMM units times 128 per unit would mean 1920 Maxwell cores.

Pretend their scheduler is great and the performance scales well with core count and clocks, and that they kept the idea of triple everything there in this hypothetical, non-existent card that is an illustration only. So we'd have a 384 bit bus with 1920 Maxwell cores, probably 7ghz memory speed of GDDR5 like Kepler does at least, and a TDP that fits inside of 200 watts. Now let's say that you only get about 75% scaling from core count here, which is reasonable even though Kepler scales pretty linearly, but it's a new architecture with Maxwell, so let's make the safe assumption. So a GTX 660 performs 12% better than a GM107 with 640 cores. Triple the core count there with our rough napkin math again with everything else and you would have a card performing around the same as GK110 fully unlocked by that theory-crafting, at least, and it has better potential for higher clocks thanks to the lower power usage.

However, in reality, we know they are going 20nm. This allows for even more power savings. This also allows for many more transistors per square mm on the die. So pretend they want to go for a 520mm2 chip on 20nm, keep costs down a tad for Big Maxwell and improve yields per wafer. According to released documents such as this: http://www.cadence.com/rl/Resources/overview/20nm_qa.pdf we can expect to see transistor counts possible of 8-12 billion. GK110 is 7.1 billion transistors. Using that as a point of reference, let's scale 28nm GK110 to 28nm Maxwell GM107: we need less memory controllers and pad space, so we can safely come up with a number in the neighborhood, considering the 148mm2 die size compared to GK110's 551mm2 size. At 20nm, you will be able to fit upwards of 11-12 billion transistors for a high-end part. For 28nm the 148mm2 die size indicates roughly 1.7-1.8 billion transistors with the 128-bit bus.

So now we have a decent number here: we know that Maxwell at 28nm in GM107 form is taking about 2 billion transistors to perform at a level about 89% as fast as a GK106. Let's use this as a base for the next part of this thought exercise smile.gif.

So 2 billion.... it's safe to say they could fit 7 GPC's at 20nm easily since 20nm should provide roughly a 2x density shrink in die size used per transistor, very easily, and we wouldn't need to duplicate memory controllers beyond 3x of the 28nm design's if we went for a 384-bit bus. At 20nm let's say they went for a 384-bit bus, to feed 7 GPC's worth of cores since the architecture is more bandwidth-efficient than Kepler, clearly by far. That'd give us 4480 Maxwell cores which already are much more bandwidth-efficient, so it probably would be plenty well-fed on that end of things.

Power-wise we'd be looking at an envelope of, 250 watts since Maxwell GM107 at 28nm takes a full-card power of 60w for 640 cores. That means probably around 45w for the GPU itself, allowing 10w for the GDDR5 and 5w for the fan and other circuitry. The move to 20nm will improve power efficiency drastically, due to the shorter gate lengths. So it's fair to say they could fit seven of those inside of that envelope, more than easily.

Again, let's go for a linear scaling factor of approximately 75% for the core count improvements... so we have a GTX 660 GK106 card we will use and compare to a GK110 780 Ti. A GTX 660 at 1080p is able to pull about 50% of the performance of a fully-unlocked, 780 Ti (source: http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_780_Ti/27.html).

Now we know each SMM will provide around 90% of that performance, and revision 2.0 (second gen) on 20nm will probably be closer to a 100% figure. So let's conservatively say each SMM-based GPC results in a real-world performance of a GTX 660. In other words, around half of a 780 Ti. Now let's conservatively also say we only see the benefit of 75% of the cores when scaling it to 7 GPC's or 35 SMM's. 35 SMM's would be, as you recall 4480 maxwell cores. Multiply the performance of the 50% card by 7 and we'd have 350% (or 2.5x faster than) of the performance of a full GK110. However, let's now apply the 75% rough rule and we come up with a much more reasonable 262.5% of the performance, or 2.6x as fast as (1.6x faster than) a GTX 780 Ti.

My predictions, therefore, are that we will see a Big Maxwell on 20nm with at minimum twice the performance of a GTX 780 Ti, and by current rumors it is due this year. Add in that Maxwell will be able to clock higher at 20nm (I based those power figures off of the numbers above which were of a card at 28nm with a 1085mhz GPU clock, and realistically they can probably fit a stock core speed of 1100-1150mhz of this big a chip in then. Thus my predictions are we would see a 450-460mm2 die size for this hypothetical card, at 20nm, with a 250-260w TDP rating and thanks to the 2x transistor density, approximately 9.2b transistors.

Napkin math, for sure, but worth thinking about, eh?


I think they will go straight to 16nm for the 2nd generation Maxwell, 20nm seems to not provide enough benefits for all the work involved in transitioning to a different process node with lower yields. Given Maxwell's efficiency at 28nm, they can make a great run of cards with none of those problems. Taking into consideration that the new process nodes are becoming more and more expensive and laborious, Nvidia engineered their way through with Maxwell and can jump to 16nm in the second half of next year in my opinion.

16nm FinFet should be ready for GPUs by then: http://www.bit-tech.net/news/hardware/2013/12/13/tsmc-16nm/1

Big gains over the GTX 780 Ti like you predict will most likely be reserved for the 2nd generation Maxwell, if anything about the GTX 600 series vs GTX 500 series is any indication (the GTX 680 was 19% faster than the GTX 580 when it launched, is currently around 26% faster), along with manufacturing processes taking longer and longer to develop.

Quote:
Originally Posted by skupples View Post

So when do you all think Denver will rear its game changing, glorious, sexy, amazing, put everyone on their asses, head? 20nm, or 16nm?

16nm is my bet, but it could be already on 28nm on the upper range cards.
Edited by tpi2007 - 3/15/14 at 10:08pm
 
Metro 2033 review
Metro 2033
CPUMotherboardGraphicsRAM
Core i7-3820 Asus Sabertooth X79 MSI GTX 750 Ti TF Gaming 16 GB Corsair DDR3 1866 Mhz Dominator 
Hard DriveOptical DriveCoolingOS
Samsung SSD 830 128GB + WD Caviar Black 1TB Sony Optiarc DVD-RW Corsair A70 + Noiseblocker M12-P Windows 7 Home Premium 64-bit 
MonitorKeyboardPowerCase
BenQ RL2455HM Cooler Master Octane Corsair AX750 Professional Modular 80 Plus Gold Cooler Master HAF 912 Plus 
Mouse
Cooler Master Octane 
  hide details  
Reply
 
Metro 2033 review
Metro 2033
CPUMotherboardGraphicsRAM
Core i7-3820 Asus Sabertooth X79 MSI GTX 750 Ti TF Gaming 16 GB Corsair DDR3 1866 Mhz Dominator 
Hard DriveOptical DriveCoolingOS
Samsung SSD 830 128GB + WD Caviar Black 1TB Sony Optiarc DVD-RW Corsair A70 + Noiseblocker M12-P Windows 7 Home Premium 64-bit 
MonitorKeyboardPowerCase
BenQ RL2455HM Cooler Master Octane Corsair AX750 Professional Modular 80 Plus Gold Cooler Master HAF 912 Plus 
Mouse
Cooler Master Octane 
  hide details  
Reply
post #928 of 1139
Quote:
Originally Posted by tpi2007 View Post


16nm FinFet should be ready for GPUs by then: http://www.bit-tech.net/news/hardware/2013/12/13/tsmc-16nm/1

Big gains over the GTX 780 Ti like you predict will most likely be reserved for the 2nd generation Maxwell, if anything about the GTX 600 series vs GTX 500 series is any indication, along with manufacturing processes taking longer and longer to develop. 16nm is my bet, but it could be already on 28nm on the upper range cards.

yeah. the first generation GFM104 would mostly be a 3072 - 3200 core Maxwell chip measuring at 300 - 330 sq mm. Nvidia would like to be conservative with die size as 20nm is the first node with dual pattern immersion litho. yield challenges are always going to be there with a bleeding edge node. The core clocks would be in the 1 Ghz range mostly due to 20nm being a power constrained node. the second generation Maxwell big die GM100 or GM200 would be a TSMC 16FF chip and would easily double GTX 780 Ti perf or go even higher
Edited by raghu78 - 3/15/14 at 10:17pm
Fragbox
(14 items)
 
  
CPUMotherboardGraphicsGraphics
Intel Core i5 2400 DH67BL AMD Radeon HD 6900 Series XFX HD 6950 2GB 
RAMRAMHard DriveOptical Drive
Corsair Corsair DDR3 1333 Mhz 2 GB Corsair DDR3 1333 Mhz 2 GB Western Digital Caviar Green SONY DVD-RW AD-7260S 
OSMonitorKeyboardPower
Windows 7 Professional 32 bit BENQ G2420HD Logitech K200 Seasonic VX550 psu 
CaseMouse
ANTEC 200 V2 Logitech mouse 
  hide details  
Reply
Fragbox
(14 items)
 
  
CPUMotherboardGraphicsGraphics
Intel Core i5 2400 DH67BL AMD Radeon HD 6900 Series XFX HD 6950 2GB 
RAMRAMHard DriveOptical Drive
Corsair Corsair DDR3 1333 Mhz 2 GB Corsair DDR3 1333 Mhz 2 GB Western Digital Caviar Green SONY DVD-RW AD-7260S 
OSMonitorKeyboardPower
Windows 7 Professional 32 bit BENQ G2420HD Logitech K200 Seasonic VX550 psu 
CaseMouse
ANTEC 200 V2 Logitech mouse 
  hide details  
Reply
post #929 of 1139
Quote:
Originally Posted by raghu78 View Post

Quote:
Originally Posted by tpi2007 View Post


16nm FinFet should be ready for GPUs by then: http://www.bit-tech.net/news/hardware/2013/12/13/tsmc-16nm/1

Big gains over the GTX 780 Ti like you predict will most likely be reserved for the 2nd generation Maxwell, if anything about the GTX 600 series vs GTX 500 series is any indication, along with manufacturing processes taking longer and longer to develop. 16nm is my bet, but it could be already on 28nm on the upper range cards.

yeah. the first generation GFM104 would mostly be a 3072 - 3200 core Maxwell chip measuring at 300 - 330 sq mm. Nvidia would like to be conservative with die size as 20nm is the first node with dual pattern immersion litho. yield challenges are always going to be there with a bleeding edge node. The core clocks would be in the 1 Ghz range mostly due to 20nm being a power constrained node. the second generation Maxwell big die GM100 or GM200 would be a TSMC 16FF chip and would easily double GTX 780 Ti perf or go even higher

I agree with what you say, except for the core count. It seems to me that with 128 Cuda cores (Maxwell) delivering 90% of the performance of 192 Cuda cores (Kepler), the core count won't have to be that high unless they want the GTX 800 series to perform substantially better than the GTX 700 series in comparison to how the GTX 600 series performed compared to the GTX 500 series. I think it will interestingly 'mimic' (relatively speaking, of course) AMD's current GCN Stream processor count (see my post above). With the GTX 750 and 750 Ti it already is.
Edited by tpi2007 - 3/15/14 at 10:32pm
 
Metro 2033 review
Metro 2033
CPUMotherboardGraphicsRAM
Core i7-3820 Asus Sabertooth X79 MSI GTX 750 Ti TF Gaming 16 GB Corsair DDR3 1866 Mhz Dominator 
Hard DriveOptical DriveCoolingOS
Samsung SSD 830 128GB + WD Caviar Black 1TB Sony Optiarc DVD-RW Corsair A70 + Noiseblocker M12-P Windows 7 Home Premium 64-bit 
MonitorKeyboardPowerCase
BenQ RL2455HM Cooler Master Octane Corsair AX750 Professional Modular 80 Plus Gold Cooler Master HAF 912 Plus 
Mouse
Cooler Master Octane 
  hide details  
Reply
 
Metro 2033 review
Metro 2033
CPUMotherboardGraphicsRAM
Core i7-3820 Asus Sabertooth X79 MSI GTX 750 Ti TF Gaming 16 GB Corsair DDR3 1866 Mhz Dominator 
Hard DriveOptical DriveCoolingOS
Samsung SSD 830 128GB + WD Caviar Black 1TB Sony Optiarc DVD-RW Corsair A70 + Noiseblocker M12-P Windows 7 Home Premium 64-bit 
MonitorKeyboardPowerCase
BenQ RL2455HM Cooler Master Octane Corsair AX750 Professional Modular 80 Plus Gold Cooler Master HAF 912 Plus 
Mouse
Cooler Master Octane 
  hide details  
Reply
post #930 of 1139
Quote:
Originally Posted by tpi2007 View Post

I agree with what you say, except for the core count. It seems to me that with 128 Cuda cores (Maxwell) delivering 90% of the performance of 192 Cuda cores (Kepler), the core count won't have to be that high unless they want the GTX 800 series to perform substantially better than the GTX 700 series in comparison to how the GTX 600 series performed compared to the GTX 500 series. I think it will interestingly mimic AMD's current GCN Stream processor count (see my post above). With the GTX 750 and 750 Ti it already is.

nVIDIA would want a card atleast 40% faster than GTX 780 Ti to give it a new gen flagship title and enthusiast pricing. I am sure Nvidia will price it at USD 700. 20nm is costly and Nvidia is going to want to maximize its margins. .
Fragbox
(14 items)
 
  
CPUMotherboardGraphicsGraphics
Intel Core i5 2400 DH67BL AMD Radeon HD 6900 Series XFX HD 6950 2GB 
RAMRAMHard DriveOptical Drive
Corsair Corsair DDR3 1333 Mhz 2 GB Corsair DDR3 1333 Mhz 2 GB Western Digital Caviar Green SONY DVD-RW AD-7260S 
OSMonitorKeyboardPower
Windows 7 Professional 32 bit BENQ G2420HD Logitech K200 Seasonic VX550 psu 
CaseMouse
ANTEC 200 V2 Logitech mouse 
  hide details  
Reply
Fragbox
(14 items)
 
  
CPUMotherboardGraphicsGraphics
Intel Core i5 2400 DH67BL AMD Radeon HD 6900 Series XFX HD 6950 2GB 
RAMRAMHard DriveOptical Drive
Corsair Corsair DDR3 1333 Mhz 2 GB Corsair DDR3 1333 Mhz 2 GB Western Digital Caviar Green SONY DVD-RW AD-7260S 
OSMonitorKeyboardPower
Windows 7 Professional 32 bit BENQ G2420HD Logitech K200 Seasonic VX550 psu 
CaseMouse
ANTEC 200 V2 Logitech mouse 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [wccftech] GeForce 800 Series Arrives in Second Half of 2014