Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [VC]GTX 1060 specifications leaked - faster than RX 480
New Posts  All Forums:Forum Nav:

[VC]GTX 1060 specifications leaked - faster than RX 480 - Page 48  

post #471 of 735
But Maxwell performed really well for when it was still NVIDIA's premier architecture. It doesn't matter that Maxwell sucks at DX12 unless you are a long tern GPU user because its already being replaced by Pascal. Its great for GCN owners like me that GCN is so great at DX12, but who is looking to buy a 380x anymore?

Its clear as day that Maxwell wasn't designed to be used in DX12 games because its a very barebones, ad-hoc architecture. NVIDIA is a R&D machine and their technology moves very fast because they can afford to be that bleeding edge, hence the more disposable and short term nature of their architectures. Meanwhile, AMD is still performing upgrades on the same GCN archtecutre that debuted almost five years ago. Kepler and Maxwell cards, when they were still in production, were superior to GCN cards in terms of perf/watt and perf/mm2.

GCN is too ahead of its time for its own good. High amounts of DX11 overhead on GCN cards is a feature, not a bug.
First Build
(10 items)
 
  
CPUMotherboardGraphicsRAM
i3-4370 ASROCK H97M PR04 380x Nitro Corsair 8GB 
Hard DriveHard DriveHard DriveOS
Crucial M500 120GB Seagage 320GB Caviar Blue 1TB Windows 8.1 x86-64 
PowerCase
XFX 550W PS07B 
  hide details  
First Build
(10 items)
 
  
CPUMotherboardGraphicsRAM
i3-4370 ASROCK H97M PR04 380x Nitro Corsair 8GB 
Hard DriveHard DriveHard DriveOS
Crucial M500 120GB Seagage 320GB Caviar Blue 1TB Windows 8.1 x86-64 
PowerCase
XFX 550W PS07B 
  hide details  
post #472 of 735
Quote:
Originally Posted by EightDee8D View Post

So, any predictions on vega (small and big) ? tongue.gif

I am not sure there is going to be a big vega.

Two vega's in nearly the same span of time considering the performance class is just cannibalization of sales. Considering the failure of Fiji sales, I don't think they ever want to do a 600mm2 die again. The just don't have the professional market for it to make the investment worth it.

I think all AMD is going to release is the 4096 shader vega card. This card should have it's double precision added back in to address the professional market. This along with the additions of Vega/polaris should make this around a 400mm2 die.

Since this card has 77% more shaders. That should lead to a straight forward 77% increase in performance. However, because utilization is more difficult is more difficult as an architecture is scaled up, scaling is negatively effected. Fiji vs full Tonga, 2x the specs but only 75-80 percent increase in performance when clocks or equalized. The same thing will happen to vega, but perhaps a bit better but there will be some scaling loss.

So if clocks are equal to Polaris, I would say performance should be around 55-65 percent better than an rx 480. Depending how the driver situation changes along with the directx 12 additions, it's competitiveness vs pascal could change. With HBM2 power savings and maturity and revisions, we might be able to get a slight clock boost vs polaris. If this happens, add 5% to these estimates.

However at the price range, we won't be getting a water cooler, so don't expect a big jump in clocks. Maybe 1300mhz or 1320mhz with around 225 watt power consumption.
post #473 of 735
Quote:
Originally Posted by MikeDuffy View Post


Maxwell is utter garbage in directx12- don't believe me? Here is a quote from SkyMtl at HardwareCanucks:

Moving on to DX12 and we see AMD’s new architecture really coming into its own against the NVIDIA cards. It absolutely demolishes the GTX 970 across the board (even in NVIDIA-friendly games like Tomb Raider) and even manages to run circles around that once-expensive GTX 980. These tests show Maxwell’s performance in current DX12 applications is nothing short of embarrassing and proves this architecture simply wasn’t designed with these types of workloads in mind.

It would be nice when someone compare it to DX12 Keppler, SMD to SMD clock per clock.

Actually, the current DX12 controversy only shows that AMD drivers utterly sucked, and only NVidia showed some competence. Well, with simple DX12 interface, the main effort has been send to game developers, which would be responsible for proper optimization. For AAA games that means just a minor change, for really small developers (including freeware), it's better to use OGL version and stay away from DX12 for anything that would need extensive optimizations. Things might be really variable on different configurations...
post #474 of 735
Could we stick with talking about GTX 1060 please ?

I know it's difficult, because one assertion leads to another and the will to discuss about it, but I think we're going far from the original thread. tongue.gif
My System
(14 items)
 
  
CPUMotherboardGraphicsRAM
i7 3770K Asrock Z77 Extreme 4 GTX670 + Arctic cooling accelero extreme II 8GB, Gskill Sniper  
Hard DriveOSMonitorPower
Western Digital/ Seagate/ Samsung Windows 10 64 bits 24 inches MVA Enermax  
CaseMouseAudio
Corsair CM Storm Xornet Soundblaster Z 
  hide details  
My System
(14 items)
 
  
CPUMotherboardGraphicsRAM
i7 3770K Asrock Z77 Extreme 4 GTX670 + Arctic cooling accelero extreme II 8GB, Gskill Sniper  
Hard DriveOSMonitorPower
Western Digital/ Seagate/ Samsung Windows 10 64 bits 24 inches MVA Enermax  
CaseMouseAudio
Corsair CM Storm Xornet Soundblaster Z 
  hide details  
post #475 of 735
Quote:
Originally Posted by MikeDuffy View Post

@Mahigan was the first to question Nvidia's claim about multi-engine performance in directx12 - the first!

He was completely correct with his analysis of Maxwell's shortcomings - funny how you guys don't seem to give him credit on that one.

Maxwell is utter garbage in directx12- don't believe me? Here is a quote from SkyMtl at HardwareCanucks:

Moving on to DX12 and we see AMD’s new architecture really coming into its own against the NVIDIA cards. It absolutely demolishes the GTX 970 across the board (even in NVIDIA-friendly games like Tomb Raider) and even manages to run circles around that once-expensive GTX 980. These tests show Maxwell’s performance in current DX12 applications is nothing short of embarrassing and proves this architecture simply wasn’t designed with these types of workloads in mind.

@tajoh111- where is Nvidia's magical async compute driver? It's been over a year now and I haven't once seen you giving them a hard time about it. Why do you guys ignore Nvidia lies?

He definitely knows his stuff but his predictions about performance are way off.

The problem for AMD is in the best case scenario of rx480 vs pascal so far is pascal being 30 percent faster than it. This isn't that much off between their difference in die size or transistor count. Meaning at equal size they almost perform the same in directx 12, but everywhere else, Nvidia has a 50% performance advantage. If Nvidia add's that directx 12 functionality, then AMD is toast.

That's why we need something better and newer than GCN, so when that day happens, AMD is not screwed.
Edited by tajoh111 - 7/6/16 at 2:41pm
post #476 of 735
Quote:
Originally Posted by DMatthewStewart View Post

How did they even come up with what htey have labeled as "power efficiency"?

Pretty simple really...typical power vs. average performance.

RX480 is ~150w 1060 is ~120w.

NVIDIA says 1060 is ~15% faster.

So, (150 / 120) * 1.15 = 1.4375
Quote:
Originally Posted by DMatthewStewart View Post

Because their graph is not even close to the numbers they report (among other things). Its almost like they were thinking "well, it did better in certain areas so we should just make the bar huge, dont worry about scale, and people wont actually compare it to our own data, so..."

I don't really have a problem with the graph. It's cut off, which may mislead some, but that's really the fault of those misreading things.
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  
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  
post #477 of 735
Quote:
Originally Posted by magnek View Post

Within 10% of 1080 performance, assuming 4096 shaders and similar clockspeeds as Polaris 10.

That's impossible to predict. Polaris and Vega don't share the same architecture. Polaris is Graphics IPv8 similar to Tonga/Fiji, while it seems Vega is Graphics IPv9 which was developed for 14nm.
post #478 of 735
@tajoh111

Rep+ for effort. thumb.gif

Don't take this the wrong way... but, all of that was/is speculation, no one knew what was Greenland/Vega, Polaris, HBM2, etc. As i said in my earlier post, it was perf per Watt that caught most people off guard, still best not to call anyone out unless they were stating everything they said as fact.

Half the fun of these forums is speculation, for me anyway. tongue.gif
post #479 of 735
Quote:
Originally Posted by variant View Post

That's impossible to predict. Polaris and Vega don't share the same architecture. Polaris is Graphics IPv8 similar to Tonga/Fiji, while it seems Vega is Graphics IPv9 which was developed for 14nm.
Better hope it's same ipv8 cuz we know what huge expectations and hype does. tongue.gif
post #480 of 735
Quote:
Originally Posted by Blameless View Post

Pretty simple really...typical power vs. average performance.

RX480 is ~150w 1060 is ~120w.

NVIDIA says 1060 is ~15% faster.

So, (150 / 120) * 1.15 = 1.4375
I don't really have a problem with the graph. It's cut off, which may mislead some, but that's really the fault of those misreading things.

Hold on a minute. That is entirely wrong. We are comparing the 120w usage to a device that uses 150w. That means that, since we are trying to equate the performance, our max amount of power used is 150w. 1060 uses 120w. That means its 20% more power effecient because its only using 80% of the total wattage calculated, or total wattage pool. 120w is 80% of 150w. So youre calculating it incorrectly just as they have done. Thats strictly full TDP. Before we even touch on performance. So 150w is 100% of the pool of power that we are rating. The 1060 uses 120w. Its 20% more power effecient

And where does 1.15 come from?

Dividing 150/120 gives us no number related to power efficiency. The original graph is incorrect because it was calculated incorrectly. Working it backwards and then typing it out seems right at first. Its not. Unless Im missing something here. Im open minded to being totally wrong but this seems way too easily identifiable as incorrect.
Edited by DMatthewStewart - 7/6/16 at 3:30pm
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Rumors and Unconfirmed Articles
This thread is locked  
Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [VC]GTX 1060 specifications leaked - faster than RX 480