Overclock.net › Forums › Industry News › Hardware News › [AdoredTV] Pascal vs Maxwell at same clockspeeds, same FLOPS (1080 vs 980 Ti)
New Posts  All Forums:Forum Nav:

[AdoredTV] Pascal vs Maxwell at same clockspeeds, same FLOPS (1080 vs 980 Ti) - Page 26

post #251 of 305
the Crazy result you linked has Strix Fury, not a ref Fury ..

Quote:
No matter. I can also cherry-pick bencmarks with a high preset;
http://www.techspot.com/review/1182-nvidia-geforce-gtx-1070/page4.html
why use an older review with older drivers ?

http://www.techspot.com/review/1209-nvidia-geforce-gtx-1060/page6.html
^ 1070 beats R9 Nano (not by much though .. i guess that makes it tie with the Fury then as you said)

so I can concede the 1070 = Fury thing ^^ .. but I was correct about Doom Vulkan results tongue.gif
Edited by ChevChelios - 7/29/16 at 8:39am
The Green Beast
(16 items)
 
  
CPUMotherboardGraphicsRAM
Intel i7 3770K @ 4500 Mhz ASRock Z77 Pro3 Gigabyte GTX 1080 G1 Gaming Crucial Ballistix 2x8GB DDR3-1600 
Hard DriveHard DriveHard DriveCooling
SSD Crucial M550 500GB SSD Samsung 850 Evo 1TB HDD Seagate 7200rpm 3TB Cooler Master Hyper 212 EVO; Xilence X5 
OSMonitorKeyboardPower
Windows 10 Pro 64-bit Acer Predator XB271HU 27" IPS Gsync 1440p 165Hz CM Storm QuickFire XT Cherry Red 800W modular 
CaseMouseMouse PadAudio
Fractal Design Define R4 Black Logitech G900 Chaos Spectrum SteelSeries QcK+ 4mm SK Gaming Realtek On-board 
  hide details  
Reply
The Green Beast
(16 items)
 
  
CPUMotherboardGraphicsRAM
Intel i7 3770K @ 4500 Mhz ASRock Z77 Pro3 Gigabyte GTX 1080 G1 Gaming Crucial Ballistix 2x8GB DDR3-1600 
Hard DriveHard DriveHard DriveCooling
SSD Crucial M550 500GB SSD Samsung 850 Evo 1TB HDD Seagate 7200rpm 3TB Cooler Master Hyper 212 EVO; Xilence X5 
OSMonitorKeyboardPower
Windows 10 Pro 64-bit Acer Predator XB271HU 27" IPS Gsync 1440p 165Hz CM Storm QuickFire XT Cherry Red 800W modular 
CaseMouseMouse PadAudio
Fractal Design Define R4 Black Logitech G900 Chaos Spectrum SteelSeries QcK+ 4mm SK Gaming Realtek On-board 
  hide details  
Reply
post #252 of 305
Quote:
Originally Posted by Pyrotagonist View Post

Hopefully Vega is built on TSMC

Highly unlikely.
Quote:
Originally Posted by EightDee8D View Post

Those high clocks and efficiency came from tsmc 16nm ff not because of Nvidia.

Not strictly true. There is a lot of interplay and collaboration between an ASIC designer and fabricator. Unless you strictly follow the fabricators design guidelines (and if you did you'd never be competitive), you have to work very closely with them to make sure your parts work as they should.

This is why you can't just switch fabricator willy-nilly...a lot of work goes in to tailoring a design for a given process.
Quote:
Originally Posted by sherlock View Post

Irrelevant when clock speed gains outstrip IPC losses/stagnation, as is the case here with 1080. If an architecture enables higher performance, whether it is done through IPC or higher clock envelope hardly matters to the consumer.

Many of of us aren't ordinary consumers. We are overclockers, tuners, speculators, and forecasters...enthusiasts really.

How performance is achieved does matter, because that how influences how much more an enthusiast can extract, under which scenarios, as well as the specific trade-offs of doing so.
Quote:
Originally Posted by junkman View Post

I think you raise a valid point. With Polaris, if 1400MHz is the limit on the 14nm FF process without large voltage adjustments, it will be interesting to see how Vega will perform.

Hopefully AMD and GloFo get whatever issues they are having straightened out as the process matures. The current process is acceptable for a mainstream part, but won't cut it for a big die that has to be competitive with GP102 at the high-end
Quote:
Originally Posted by junkman View Post

With that concern, I wonder if HBM2 memory architecture power consumption savings will be enough to enable a higher voltage threshold for Vega.

Doesn't really work like that.

As I stated earlier, consumer parts already tend to be clocked far beyond the point of diminishing returns with respect to performance/watt. Saving a dozen or two watts with HBM and a better tuned board won't allow the GPU power budget to be increased enough to really matter, not without some significant improvement to process yields at the same time.
Edited by Blameless - 7/29/16 at 8:38am
Primary
(15 items)
 
Secondary
(13 items)
 
In progress
(10 items)
 
CPUMotherboardGraphicsRAM
5820K @ 4.2/3.5GHz core/uncore, 1.175/1.15v Gigabyte X99 SOC Champion (F22n) Gigabyte AORUS GTX 1080 Ti (F3P) @ 2025/1485, 1... 4x4GiB Crucial @ 2667, 12-12-12-28-T1, 1.34v 
Hard DriveHard DriveHard DriveCooling
Plextor M6e 128GB (fw 1.06) M.2 (PCI-E 2.0 2x) 2x Crucial M4 256GB 4x WD Scorpio Black 500GB Noctua NH-D15 
OSMonitorKeyboardPower
Windows 7 Professional x64 SP1 BenQ BL3200PT Filco Majestouch Tenkeyless (MX Brown) Corsair RM1000x 
CaseMouseAudio
Fractal Design Define R4 Logitech G402 Realtek ALC1150 + M-Audio AV40 
CPUMotherboardGraphicsRAM
X5670 @ 4.4/3.2GHz core/uncore, 1.36 vcore, 1.2... Gigabyte X58A-UD5 r2.0 w/FF3mod10 BIOS Sapphire Fury Nitro OC+ @ 1053/500, 1.225vGPU/1... 2x Samsung MV-3V4G3D/US @ 2000, 10-11-11-30-T1,... 
RAMHard DriveHard DriveHard Drive
1x Crucial BLT4G3D1608ET3LX0 @ 2000, 10-11-11-3... OCZ (Toshiba) Trion 150 120GB Hyundai Sapphire 120GB 3x Hitachi Deskstar 7k1000.C 1TB 
CoolingOSPowerCase
Noctua NH-D14 Windows 7 Pro x64 SP1 Antec TP-750 Fractal Design R5 
Audio
ASUS Xonar DS 
CPUMotherboardGraphicsRAM
i7-6800K @ 4.3/3.5GHz core/uncore, 1.36/1.2v ASRock X99 OC Formula (P3.10) GTX 780 (temporary) 4x4GiB Crucial DDR4-2400 @ 11-13-12-28-T2, 1.33v 
Hard DriveHard DriveCoolingOS
Intel 600p 256GB NVMe 2x HGST Travelstar 7k1000 1TB Corsair H55 (temporary) Windows Server 2016 Datacenter 
PowerCase
Seasonic SS-860XP2 Corsair Carbide Air 540 
  hide details  
Reply
Primary
(15 items)
 
Secondary
(13 items)
 
In progress
(10 items)
 
CPUMotherboardGraphicsRAM
5820K @ 4.2/3.5GHz core/uncore, 1.175/1.15v Gigabyte X99 SOC Champion (F22n) Gigabyte AORUS GTX 1080 Ti (F3P) @ 2025/1485, 1... 4x4GiB Crucial @ 2667, 12-12-12-28-T1, 1.34v 
Hard DriveHard DriveHard DriveCooling
Plextor M6e 128GB (fw 1.06) M.2 (PCI-E 2.0 2x) 2x Crucial M4 256GB 4x WD Scorpio Black 500GB Noctua NH-D15 
OSMonitorKeyboardPower
Windows 7 Professional x64 SP1 BenQ BL3200PT Filco Majestouch Tenkeyless (MX Brown) Corsair RM1000x 
CaseMouseAudio
Fractal Design Define R4 Logitech G402 Realtek ALC1150 + M-Audio AV40 
CPUMotherboardGraphicsRAM
X5670 @ 4.4/3.2GHz core/uncore, 1.36 vcore, 1.2... Gigabyte X58A-UD5 r2.0 w/FF3mod10 BIOS Sapphire Fury Nitro OC+ @ 1053/500, 1.225vGPU/1... 2x Samsung MV-3V4G3D/US @ 2000, 10-11-11-30-T1,... 
RAMHard DriveHard DriveHard Drive
1x Crucial BLT4G3D1608ET3LX0 @ 2000, 10-11-11-3... OCZ (Toshiba) Trion 150 120GB Hyundai Sapphire 120GB 3x Hitachi Deskstar 7k1000.C 1TB 
CoolingOSPowerCase
Noctua NH-D14 Windows 7 Pro x64 SP1 Antec TP-750 Fractal Design R5 
Audio
ASUS Xonar DS 
CPUMotherboardGraphicsRAM
i7-6800K @ 4.3/3.5GHz core/uncore, 1.36/1.2v ASRock X99 OC Formula (P3.10) GTX 780 (temporary) 4x4GiB Crucial DDR4-2400 @ 11-13-12-28-T2, 1.33v 
Hard DriveHard DriveCoolingOS
Intel 600p 256GB NVMe 2x HGST Travelstar 7k1000 1TB Corsair H55 (temporary) Windows Server 2016 Datacenter 
PowerCase
Seasonic SS-860XP2 Corsair Carbide Air 540 
  hide details  
Reply
post #253 of 305
Quote:
Originally Posted by ChevChelios View Post

the Crazy result you linked has Strix Fury, not a ref Fury ..
why use an older review with older drivers ?

http://www.techspot.com/review/1209-nvidia-geforce-gtx-1060/page6.html
^ 1070 beats R9 Nano (not by much though .. i guess that makes it tie with the Fury then as you said)
a, there is no reference fury, b, ashes shouldnt be used as a dx12 comparison as it heavily favors nvidia wink.gif
post #254 of 305
Quote:
Originally Posted by ChevChelios View Post

the Crazy result you linked has Strix Fury, not a ref Fury
There never was a reference Fury, only partner cards from Asus, Gigabyte and Sapphire.
post #255 of 305
alright, I already edited my post about him being correct about the 1070 = Fury in Ashes

Quote:
ashes shouldnt be used as a dx12 comparison as it heavily favors nvidia
mind blown
The Green Beast
(16 items)
 
  
CPUMotherboardGraphicsRAM
Intel i7 3770K @ 4500 Mhz ASRock Z77 Pro3 Gigabyte GTX 1080 G1 Gaming Crucial Ballistix 2x8GB DDR3-1600 
Hard DriveHard DriveHard DriveCooling
SSD Crucial M550 500GB SSD Samsung 850 Evo 1TB HDD Seagate 7200rpm 3TB Cooler Master Hyper 212 EVO; Xilence X5 
OSMonitorKeyboardPower
Windows 10 Pro 64-bit Acer Predator XB271HU 27" IPS Gsync 1440p 165Hz CM Storm QuickFire XT Cherry Red 800W modular 
CaseMouseMouse PadAudio
Fractal Design Define R4 Black Logitech G900 Chaos Spectrum SteelSeries QcK+ 4mm SK Gaming Realtek On-board 
  hide details  
Reply
The Green Beast
(16 items)
 
  
CPUMotherboardGraphicsRAM
Intel i7 3770K @ 4500 Mhz ASRock Z77 Pro3 Gigabyte GTX 1080 G1 Gaming Crucial Ballistix 2x8GB DDR3-1600 
Hard DriveHard DriveHard DriveCooling
SSD Crucial M550 500GB SSD Samsung 850 Evo 1TB HDD Seagate 7200rpm 3TB Cooler Master Hyper 212 EVO; Xilence X5 
OSMonitorKeyboardPower
Windows 10 Pro 64-bit Acer Predator XB271HU 27" IPS Gsync 1440p 165Hz CM Storm QuickFire XT Cherry Red 800W modular 
CaseMouseMouse PadAudio
Fractal Design Define R4 Black Logitech G900 Chaos Spectrum SteelSeries QcK+ 4mm SK Gaming Realtek On-board 
  hide details  
Reply
post #256 of 305
The talk is always about Ashes, Hitman, a bunch of GameWorks games like Witcher 3, Fallout 4, Rise of the Tomb Raider... The two extremes...

For some reason, no one mentions Call of Duty Black Ops III, where AMD literally wipes the floor with nVidia. As in, R9 390X matches 980 Ti.
Edited by NightAntilli - 7/29/16 at 8:54am
post #257 of 305
Quote:
Originally Posted by NightAntilli View Post

The talk is always about Ashes, Hitman, a bunch of GameWorks games like Witcher 3, Fallout 4, Rise of the Tomb Raider... The two extremes...

For some reason, no one mentions Call of Duty Black Ops III, where AMD literally wipes the floor with nVidia. As in, R9 390X matches 980 Ti.

I just want to see some Elite: Dangerous Horizons benchmarks.
Primary
(15 items)
 
Secondary
(13 items)
 
In progress
(10 items)
 
CPUMotherboardGraphicsRAM
5820K @ 4.2/3.5GHz core/uncore, 1.175/1.15v Gigabyte X99 SOC Champion (F22n) Gigabyte AORUS GTX 1080 Ti (F3P) @ 2025/1485, 1... 4x4GiB Crucial @ 2667, 12-12-12-28-T1, 1.34v 
Hard DriveHard DriveHard DriveCooling
Plextor M6e 128GB (fw 1.06) M.2 (PCI-E 2.0 2x) 2x Crucial M4 256GB 4x WD Scorpio Black 500GB Noctua NH-D15 
OSMonitorKeyboardPower
Windows 7 Professional x64 SP1 BenQ BL3200PT Filco Majestouch Tenkeyless (MX Brown) Corsair RM1000x 
CaseMouseAudio
Fractal Design Define R4 Logitech G402 Realtek ALC1150 + M-Audio AV40 
CPUMotherboardGraphicsRAM
X5670 @ 4.4/3.2GHz core/uncore, 1.36 vcore, 1.2... Gigabyte X58A-UD5 r2.0 w/FF3mod10 BIOS Sapphire Fury Nitro OC+ @ 1053/500, 1.225vGPU/1... 2x Samsung MV-3V4G3D/US @ 2000, 10-11-11-30-T1,... 
RAMHard DriveHard DriveHard Drive
1x Crucial BLT4G3D1608ET3LX0 @ 2000, 10-11-11-3... OCZ (Toshiba) Trion 150 120GB Hyundai Sapphire 120GB 3x Hitachi Deskstar 7k1000.C 1TB 
CoolingOSPowerCase
Noctua NH-D14 Windows 7 Pro x64 SP1 Antec TP-750 Fractal Design R5 
Audio
ASUS Xonar DS 
CPUMotherboardGraphicsRAM
i7-6800K @ 4.3/3.5GHz core/uncore, 1.36/1.2v ASRock X99 OC Formula (P3.10) GTX 780 (temporary) 4x4GiB Crucial DDR4-2400 @ 11-13-12-28-T2, 1.33v 
Hard DriveHard DriveCoolingOS
Intel 600p 256GB NVMe 2x HGST Travelstar 7k1000 1TB Corsair H55 (temporary) Windows Server 2016 Datacenter 
PowerCase
Seasonic SS-860XP2 Corsair Carbide Air 540 
  hide details  
Reply
Primary
(15 items)
 
Secondary
(13 items)
 
In progress
(10 items)
 
CPUMotherboardGraphicsRAM
5820K @ 4.2/3.5GHz core/uncore, 1.175/1.15v Gigabyte X99 SOC Champion (F22n) Gigabyte AORUS GTX 1080 Ti (F3P) @ 2025/1485, 1... 4x4GiB Crucial @ 2667, 12-12-12-28-T1, 1.34v 
Hard DriveHard DriveHard DriveCooling
Plextor M6e 128GB (fw 1.06) M.2 (PCI-E 2.0 2x) 2x Crucial M4 256GB 4x WD Scorpio Black 500GB Noctua NH-D15 
OSMonitorKeyboardPower
Windows 7 Professional x64 SP1 BenQ BL3200PT Filco Majestouch Tenkeyless (MX Brown) Corsair RM1000x 
CaseMouseAudio
Fractal Design Define R4 Logitech G402 Realtek ALC1150 + M-Audio AV40 
CPUMotherboardGraphicsRAM
X5670 @ 4.4/3.2GHz core/uncore, 1.36 vcore, 1.2... Gigabyte X58A-UD5 r2.0 w/FF3mod10 BIOS Sapphire Fury Nitro OC+ @ 1053/500, 1.225vGPU/1... 2x Samsung MV-3V4G3D/US @ 2000, 10-11-11-30-T1,... 
RAMHard DriveHard DriveHard Drive
1x Crucial BLT4G3D1608ET3LX0 @ 2000, 10-11-11-3... OCZ (Toshiba) Trion 150 120GB Hyundai Sapphire 120GB 3x Hitachi Deskstar 7k1000.C 1TB 
CoolingOSPowerCase
Noctua NH-D14 Windows 7 Pro x64 SP1 Antec TP-750 Fractal Design R5 
Audio
ASUS Xonar DS 
CPUMotherboardGraphicsRAM
i7-6800K @ 4.3/3.5GHz core/uncore, 1.36/1.2v ASRock X99 OC Formula (P3.10) GTX 780 (temporary) 4x4GiB Crucial DDR4-2400 @ 11-13-12-28-T2, 1.33v 
Hard DriveHard DriveCoolingOS
Intel 600p 256GB NVMe 2x HGST Travelstar 7k1000 1TB Corsair H55 (temporary) Windows Server 2016 Datacenter 
PowerCase
Seasonic SS-860XP2 Corsair Carbide Air 540 
  hide details  
Reply
post #258 of 305
Quote:
Originally Posted by Blameless View Post

An ideal comparison would be same number of functional units, ROPs, TMUs, shaders, memory channels, at identical, fixed clocks. However, there are no Maxwell vs. Pascal parts that align so perfectly.


Running slightly different SP counts at slightly different core speeds (what was done in the last portion of testing) is close enough for the sake of argument when comparing the shaders in and of themselves. As a rough demonstration of the relative lack of core architectural differences, a basic FLOP for FLOP comparison illustrates things well enough.
980 vs. 1080 would be a much higher level comparison than what this was going for. The 1080 has more of everything (as is a given) and it would be even more difficult to get the theoretical shader performance of the two parts to align.
Pick three?

This.

Had it not been bottlenecked, otherwise, the Fury X with its 4096 shaders would have eaten the Titan X for breakfast. Sadly, the Fury X was bottllenecked and needed more triangle output, along with RBEs (which contain color and Z/Stencil ROPs). It had only 16 RBEs, so 64 color ROPs and 256 Z/Stencil ROPs. Triangle output was a big problem as well.

Hopefully if there is a 4096 core Vega, we see at least 24 RBEs (96 color ROPs and 384 Z/Stencil ROPs) and ideally 32 RBEs (128 color ROPs and 512 Z/Stencil ROPs). The 6144 core part would need 50% more again at 48 RBEs (for a total of 192 color ROPs and 768 Z/Stencil RBEs). The 4096 part will need double the triangle output of the 2304 part Polaris and the 6144 core part triple. They need more than 4 shader engines - ideally 8 and 12 would work best IMO.

The GTX 1080 and the 980Ti are close enough though. One is 2560 shaders and the other is 2816 shaders, which is 10%. That said, it is looking like the main advantage has been the 16/20nm process, not any architectural refinements in Maxwell, at least at DX11. DX12 may help Pascal a bit more, but without anything like the Command Processor + 8 ACEs on AMD's GPUs, it is not truly parallel, so it remains at a drawback there.

Edit:
I should note that there have been some advances in Pascal over Maxwell, most notably the Delta Color Compression improvements. Those have not translated into major game fps improvements though, suggesting that the memory bandwidth is not the bottleneck.

Quote:
Originally Posted by Blameless View Post

Warning: Spoiler! (Click to show)
For much of what I do, a GPU is just a collection of narrow, very parallel, CPUs.
IPC comparisons are precisely the sort of thing can help dissect a pipeline, in the absence of detailed technical specifications. This becomes more evident the broader the spectrum of software you run on something as IPCs are going to vary from task to task in different architectures. The absence of such variance strongly implies a very similar architecture.
Knowing the relative IPC of my VLIW5 5800 series parts vs. my VLIW4 6900 series parts (turns out that, shader for shader, they were almost exactly same in SHA256 and scrypt hashing, all other things being equal) was of critical importance to deciding which ones were most worthwhile to include in my mining farm (which ever ones could clock their memory the highest for scrypt, and whichever one's had the best core MHz/watt for SHA256) and which ones to retire (everything else). If I had made an incorrect assumption about IPC, I would have had a less profitable farm.

Understanding architectural differences is also useful in forecasting future performance of unreleased parts. Once you know how clock speed and unit counts affect performance, it becomes pretty easy to ballpark how something of the same or similar architecture will perform, well before anyone has test samples.

As for clock speeds, they are just one aspect of performance, IPC is another. You can't predict performance in the absence of one. I may know a part is going to be, or can be, clocked 30% higher, but the less certain I am about architectural differences, the more in the dark I am about final performance.
No it's not, which is why I thought his comments about the 980 Ti being faster in some tests despite the same FLOPs was pretty ignorant and why I would likely to have seen some purely shader limited tests (admittedly hard to find in games).

Still, there was enough of a selection to get a rough idea and that does reinforce what was already known (that, at least at the shader level, Maxwell and Pascal look pretty similar).
Hard to say as clock potential depends on process as well as architecture.


The Beyond 3D benchmarks may be your best bet in terms of finding shader performance.




it does seem like the higher clocks account for most of the improvements.

Quote:
Originally Posted by Blameless View Post


If we extrapolate from the difference between AMD and NVIDIA peak clock trends at 28nm and assume as similar relative gap at TSMC 16nm, we might have had a Polaris 10 (RX480) that could have been pushed into the ~1.7GHz range. Just a rough guesstimate though.

Still hope for GloFo...Polaris 10 is the first big part built on it, and yields may well improve to the point that an appreciable clock jump is possible before long. TSMC did have a significant head start though.

It is entirely possible that the Samsung/GF 14/20nm process simply isn't good. Certainly the Samsung version of the Apple A8 doesn't give much room for confidence, the TSMC version consumed less power.

A GF version of that process would be even less mature and may have even worse speeds. Compounding the problem, GF has no real experience with discrete GPUs. Ok, they build Carrizo APUs, but that is it.

Quote:
Originally Posted by Blameless View Post

Highly unlikely.

I wouldn't be so quick to discount TSMC making Vega. There are rumors going around that may be possible.

I actually think that it is plausible. GF has no experience making giant reticle limit sized chips. I'd imagine TSMC would be far better for this. They have years of experience on this.

Outside of OCN, I have heard rumors that they might use TSMC for the big Vega.
Edited by CrazyElf - 7/29/16 at 10:55am
Trooper Typhoon
(20 items)
 
  
CPUMotherboardGraphicsGraphics
5960X X99A Godlike MSI 1080 Ti Lightning MSI 1080 Ti Lightning 
RAMHard DriveHard DriveHard Drive
G.Skill Trident Z 32 Gb Samsung 850 Pro Samsung SM843T 960 GB Western Digital Caviar Black 2Tb 
Hard DriveOptical DriveCoolingCooling
Samsung SV843 960 GB LG WH14NS40 Cryorig R1 Ultimate 9x Gentle Typhoon 1850 rpm on case 
OSMonitorKeyboardPower
Windows 7 Pro x64 LG 27UD68 Ducky Legend with Vortex PBT Doubleshot Backlit... EVGA 1300W G2 
CaseMouseAudioOther
Cooler Master Storm Trooper Logitech G502 Proteus Asus Xonar Essence STX Lamptron Fanatic Fan Controller  
  hide details  
Reply
Trooper Typhoon
(20 items)
 
  
CPUMotherboardGraphicsGraphics
5960X X99A Godlike MSI 1080 Ti Lightning MSI 1080 Ti Lightning 
RAMHard DriveHard DriveHard Drive
G.Skill Trident Z 32 Gb Samsung 850 Pro Samsung SM843T 960 GB Western Digital Caviar Black 2Tb 
Hard DriveOptical DriveCoolingCooling
Samsung SV843 960 GB LG WH14NS40 Cryorig R1 Ultimate 9x Gentle Typhoon 1850 rpm on case 
OSMonitorKeyboardPower
Windows 7 Pro x64 LG 27UD68 Ducky Legend with Vortex PBT Doubleshot Backlit... EVGA 1300W G2 
CaseMouseAudioOther
Cooler Master Storm Trooper Logitech G502 Proteus Asus Xonar Essence STX Lamptron Fanatic Fan Controller  
  hide details  
Reply
post #259 of 305
Quote:
Originally Posted by CrazyElf View Post

I wouldn't be so quick to discount TSMC making Vega. There are rumors going around that may be possible.

I actually think that it is plausible.

Any links to such rumors?
Quote:
Originally Posted by CrazyElf View Post

GF has no experience making giant reticle limit sized chips.

GF didn't have any experience with 14nm volume production chips until Polaris.
Primary
(15 items)
 
Secondary
(13 items)
 
In progress
(10 items)
 
CPUMotherboardGraphicsRAM
5820K @ 4.2/3.5GHz core/uncore, 1.175/1.15v Gigabyte X99 SOC Champion (F22n) Gigabyte AORUS GTX 1080 Ti (F3P) @ 2025/1485, 1... 4x4GiB Crucial @ 2667, 12-12-12-28-T1, 1.34v 
Hard DriveHard DriveHard DriveCooling
Plextor M6e 128GB (fw 1.06) M.2 (PCI-E 2.0 2x) 2x Crucial M4 256GB 4x WD Scorpio Black 500GB Noctua NH-D15 
OSMonitorKeyboardPower
Windows 7 Professional x64 SP1 BenQ BL3200PT Filco Majestouch Tenkeyless (MX Brown) Corsair RM1000x 
CaseMouseAudio
Fractal Design Define R4 Logitech G402 Realtek ALC1150 + M-Audio AV40 
CPUMotherboardGraphicsRAM
X5670 @ 4.4/3.2GHz core/uncore, 1.36 vcore, 1.2... Gigabyte X58A-UD5 r2.0 w/FF3mod10 BIOS Sapphire Fury Nitro OC+ @ 1053/500, 1.225vGPU/1... 2x Samsung MV-3V4G3D/US @ 2000, 10-11-11-30-T1,... 
RAMHard DriveHard DriveHard Drive
1x Crucial BLT4G3D1608ET3LX0 @ 2000, 10-11-11-3... OCZ (Toshiba) Trion 150 120GB Hyundai Sapphire 120GB 3x Hitachi Deskstar 7k1000.C 1TB 
CoolingOSPowerCase
Noctua NH-D14 Windows 7 Pro x64 SP1 Antec TP-750 Fractal Design R5 
Audio
ASUS Xonar DS 
CPUMotherboardGraphicsRAM
i7-6800K @ 4.3/3.5GHz core/uncore, 1.36/1.2v ASRock X99 OC Formula (P3.10) GTX 780 (temporary) 4x4GiB Crucial DDR4-2400 @ 11-13-12-28-T2, 1.33v 
Hard DriveHard DriveCoolingOS
Intel 600p 256GB NVMe 2x HGST Travelstar 7k1000 1TB Corsair H55 (temporary) Windows Server 2016 Datacenter 
PowerCase
Seasonic SS-860XP2 Corsair Carbide Air 540 
  hide details  
Reply
Primary
(15 items)
 
Secondary
(13 items)
 
In progress
(10 items)
 
CPUMotherboardGraphicsRAM
5820K @ 4.2/3.5GHz core/uncore, 1.175/1.15v Gigabyte X99 SOC Champion (F22n) Gigabyte AORUS GTX 1080 Ti (F3P) @ 2025/1485, 1... 4x4GiB Crucial @ 2667, 12-12-12-28-T1, 1.34v 
Hard DriveHard DriveHard DriveCooling
Plextor M6e 128GB (fw 1.06) M.2 (PCI-E 2.0 2x) 2x Crucial M4 256GB 4x WD Scorpio Black 500GB Noctua NH-D15 
OSMonitorKeyboardPower
Windows 7 Professional x64 SP1 BenQ BL3200PT Filco Majestouch Tenkeyless (MX Brown) Corsair RM1000x 
CaseMouseAudio
Fractal Design Define R4 Logitech G402 Realtek ALC1150 + M-Audio AV40 
CPUMotherboardGraphicsRAM
X5670 @ 4.4/3.2GHz core/uncore, 1.36 vcore, 1.2... Gigabyte X58A-UD5 r2.0 w/FF3mod10 BIOS Sapphire Fury Nitro OC+ @ 1053/500, 1.225vGPU/1... 2x Samsung MV-3V4G3D/US @ 2000, 10-11-11-30-T1,... 
RAMHard DriveHard DriveHard Drive
1x Crucial BLT4G3D1608ET3LX0 @ 2000, 10-11-11-3... OCZ (Toshiba) Trion 150 120GB Hyundai Sapphire 120GB 3x Hitachi Deskstar 7k1000.C 1TB 
CoolingOSPowerCase
Noctua NH-D14 Windows 7 Pro x64 SP1 Antec TP-750 Fractal Design R5 
Audio
ASUS Xonar DS 
CPUMotherboardGraphicsRAM
i7-6800K @ 4.3/3.5GHz core/uncore, 1.36/1.2v ASRock X99 OC Formula (P3.10) GTX 780 (temporary) 4x4GiB Crucial DDR4-2400 @ 11-13-12-28-T2, 1.33v 
Hard DriveHard DriveCoolingOS
Intel 600p 256GB NVMe 2x HGST Travelstar 7k1000 1TB Corsair H55 (temporary) Windows Server 2016 Datacenter 
PowerCase
Seasonic SS-860XP2 Corsair Carbide Air 540 
  hide details  
Reply
post #260 of 305
Quote:
Originally Posted by Blameless View Post

Any links to such rumors?
GF didn't have any experience with 14nm volume production chips until Polaris.

Here is an example off the top of my head. Will check later - I think there was some discussion and an article on 3D Center about it: (edit: Typo)

http://www.extremetech.com/computing/232493-amd-has-built-hardware-at-samsung-could-tap-foundry-for-future-products
Quote:
As of this writing and to the best of our knowledge, AMD hasn’t made any major changes to where it builds its current hardware. Its 28nm GPUs, including Fiji, are TSMC parts. Polaris 10 and 11 are built at GlobalFoundries. AMD’s 28nm APUs and 32nm Piledriver products are all GloFo parts as well, while its 28nm Jaguar and Puma-derived APUs were initially TSMC parts, but the socketed version of these chips were built at GloFo. It is assumed that Zen will be a GloFo part on 14nm and we don’t know if Vega is built at TSMC or GlobalFoundries. I’ve theorized that it’ll be a TSMC chip, since GloFo’s 14nm process isn’t tuned to build GPUs with 200W – 250W TDPs, but that’s purely a theory on my part.

I suppose we should also then be speculating about whether or not we could even see AMD parts made at Samsung, if Joel is correct.

Strictly from an engineering POV though, I can see why TSMC would make sense. If a 2304 SP part is 232mm^2, a 4096 Vega Part would be perhaps 400mm^2 (unless they find a way to pack the transistors extremely densely, which isn't out of the question, as that happened between the 7970 and the 290X). Then a giant 6144 core part would be 600mm^2. I can see why it'd be extremely risky to put that on GF.

Edit:
One other question I have is why would they secure fab space at Samsung? Is there something going on with GF that we don't know about?
Edited by CrazyElf - 7/29/16 at 10:01am
Trooper Typhoon
(20 items)
 
  
CPUMotherboardGraphicsGraphics
5960X X99A Godlike MSI 1080 Ti Lightning MSI 1080 Ti Lightning 
RAMHard DriveHard DriveHard Drive
G.Skill Trident Z 32 Gb Samsung 850 Pro Samsung SM843T 960 GB Western Digital Caviar Black 2Tb 
Hard DriveOptical DriveCoolingCooling
Samsung SV843 960 GB LG WH14NS40 Cryorig R1 Ultimate 9x Gentle Typhoon 1850 rpm on case 
OSMonitorKeyboardPower
Windows 7 Pro x64 LG 27UD68 Ducky Legend with Vortex PBT Doubleshot Backlit... EVGA 1300W G2 
CaseMouseAudioOther
Cooler Master Storm Trooper Logitech G502 Proteus Asus Xonar Essence STX Lamptron Fanatic Fan Controller  
  hide details  
Reply
Trooper Typhoon
(20 items)
 
  
CPUMotherboardGraphicsGraphics
5960X X99A Godlike MSI 1080 Ti Lightning MSI 1080 Ti Lightning 
RAMHard DriveHard DriveHard Drive
G.Skill Trident Z 32 Gb Samsung 850 Pro Samsung SM843T 960 GB Western Digital Caviar Black 2Tb 
Hard DriveOptical DriveCoolingCooling
Samsung SV843 960 GB LG WH14NS40 Cryorig R1 Ultimate 9x Gentle Typhoon 1850 rpm on case 
OSMonitorKeyboardPower
Windows 7 Pro x64 LG 27UD68 Ducky Legend with Vortex PBT Doubleshot Backlit... EVGA 1300W G2 
CaseMouseAudioOther
Cooler Master Storm Trooper Logitech G502 Proteus Asus Xonar Essence STX Lamptron Fanatic Fan Controller  
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [AdoredTV] Pascal vs Maxwell at same clockspeeds, same FLOPS (1080 vs 980 Ti)