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 46  

post #451 of 735
I can't believe what I'm seeing, are people really okay with paying over $250 for a xx60 model?!!!?? If it's 15% faster than the rx480 and costs more than $250 it's a failure in my books. People were crying about how the rx480 was disappointing dor having only 970 levels of perf well this'll be just as much of a disappointment if it comes in at the rumored price.
Mike Wazowski
(8 items)
 
  
Mike Wazowski
(8 items)
 
  
post #452 of 735
Quote:
Originally Posted by jellybeans69 View Post

Judging by suppliers 300$ in murica and 300-360e in europe. No official conf. yet though
The same old GTX 970 performance with the same old GTX 970 price. Pascal has been such an epic fail price-wise.
BEEST
(21 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i7 3770 Gigabyte Z77X-UD3H EVGA Geforce GTX 1070 SC Gaming Corsair 16GB (4x4) Vengeance LP DDR3 
Hard DriveHard DriveHard DriveHard Drive
Crucial M4 128GB Western Digital WD20EARX Western Digital WD40EFRX Western Digital WD30EZRX 
Hard DriveOptical DriveCoolingOS
Samsung 850 EVO 500GB Sony DVD-RW DRU-190S Noctua NH-U12P SE2 Windows 7 Premium 64-bit 
MonitorKeyboardPowerCase
Samsung SyncMaster B2430L Logitech DeLuxe 250 Seasonic X-660 Fractal Design R4 Black Pearl 
MouseMouse PadAudioAudio
Logitech G203 Prodigy SteelSeries 4HD Creative X-Fi Titanium Edifier C2XD 
Other
xbox360 controller 
  hide details  
BEEST
(21 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i7 3770 Gigabyte Z77X-UD3H EVGA Geforce GTX 1070 SC Gaming Corsair 16GB (4x4) Vengeance LP DDR3 
Hard DriveHard DriveHard DriveHard Drive
Crucial M4 128GB Western Digital WD20EARX Western Digital WD40EFRX Western Digital WD30EZRX 
Hard DriveOptical DriveCoolingOS
Samsung 850 EVO 500GB Sony DVD-RW DRU-190S Noctua NH-U12P SE2 Windows 7 Premium 64-bit 
MonitorKeyboardPowerCase
Samsung SyncMaster B2430L Logitech DeLuxe 250 Seasonic X-660 Fractal Design R4 Black Pearl 
MouseMouse PadAudioAudio
Logitech G203 Prodigy SteelSeries 4HD Creative X-Fi Titanium Edifier C2XD 
Other
xbox360 controller 
  hide details  
post #453 of 735
Quote:
Originally Posted by Roadkill95 View Post

I can't believe what I'm seeing, are people really okay with paying over $250 for a xx60 model?!!!?? If it's 15% faster than the rx480 and costs more than $250 it's a failure in my books. People were crying about how the rx480 was disappointing dor having only 970 levels of perf well this'll be just as much of a disappointment if it comes in at the rumored price.

Ye but nvidia is not touting the card to be a "revolution" and people are not hyping it up to match a 980ti after oc, major difference.

It will have its place, for people that dont want to put up with the awful reference model of the 480 and for people that want g-sync(a game changer imo, you dont need it but once you try it it's hard to go back).
post #454 of 735
Quote:
Originally Posted by sugalumps View Post

Ye but nvidia is not touting the card to be a "revolution" and people are not hyping it up to match a 980ti after oc, major difference.

It will have its place, for people that dont want to put up with the awful reference model of the 480 and for people that want g-sync(a game changer imo, you dont need it but once you try it it's hard to go back).
"G-Sync" with a GTX1060? you serious? the monitor is way more expensive than the card, its not even funny that you could afford a decent 144Hz monitor for half the price.

if you could afford a G-Sync monitor, you'd obviously can afford a GTX1070, better yet GTX1080 for a more consistent FPS at max settings.
post #455 of 735
The more i follow this, the more nonsense i read, incredible
post #456 of 735
There is a VERY important bar missing from that fake graph! Can we see a support timeline for the 1060 or is it too short to graph? biggrin.gif


So much speculation is being thrown around and then people regurgitate it as fact, because, well I read it on the internet so it must be true! . . . .
Tiny Big
(9 items)
 
  
CPUMotherboardGraphicsRAM
R7 1700 ASRock AB350 Gaming-ITX Red Devil RX580 8GB Patriot Viper 4 3200mhz 16GB (2x8GB) 
Hard DriveHard DriveOSPower
Crucial MX300 M.2 2280 (275GB) Samsung F3 (1TB) Win7 Home Corsair CX550M (550w) 
Case
Silverstone SG13B-Q (ITX) 
  hide details  
Tiny Big
(9 items)
 
  
CPUMotherboardGraphicsRAM
R7 1700 ASRock AB350 Gaming-ITX Red Devil RX580 8GB Patriot Viper 4 3200mhz 16GB (2x8GB) 
Hard DriveHard DriveOSPower
Crucial MX300 M.2 2280 (275GB) Samsung F3 (1TB) Win7 Home Corsair CX550M (550w) 
Case
Silverstone SG13B-Q (ITX) 
  hide details  
post #457 of 735
Quote:
Originally Posted by SlackerITGuy View Post

This launch should be somewhat exciting.

But first NVIDIA needs to solve the horrible DPC Latancy issues, micro-stuttering and low performance that's affecting a lot of Pascal buyers.
Wow, thats still going on? I recall that as far back as the gtx480.
I thought that was fixed doh.gif

Quote:
Originally Posted by Clocknut View Post

dont really care about power consumption. The price is really what I care.

I hope is not above $250.

If it is 120w-130w TDP, I might be able to run SLI within 650w spec. (buy 1 now, buy another 1-2yr later)
If they release a 1060 for $199 I will be honestly shocked. Good ole competition would actually be...working? poke.gif
Quote:
Originally Posted by andrews2547 View Post

I am completely serious.

Who's to say they didn't only test content that only uses Gameworks or DX11 to get those results? Or if they used one game to get those results. I'm not denying the GTX 1060 is going to be more powerful than the RX480. I'm just pointing out the graph that is in the source is terrible. It's worse than the graphs that AMD used to show 2x RX480s are "better" than a single GTX 1080. Those graphs at least said how they got the results.
I'm not denying the GTX 1060 is more power efficient. I just want to know what tests were done to get that specific results. If they just said "The GTX 1060 is more power efficient than the RX480" then it wouldn't be an issue, but they figures for it, which means tests were done but they aren't revealing what tests were done.

This is assuming that graph isn't made up by Videocardz to get ad revenue.
By now I simply assume they used games heavy in GW to get any of their performance results. I suspect a batman game to be in there somewhere.
post #458 of 735
Quote:
Originally Posted by epic1337 View Post

"G-Sync" with a GTX1060? you serious? the monitor is way more expensive than the card, its not even funny that you could afford a decent 144Hz monitor for half the price.

if you could afford a G-Sync monitor, you'd obviously can afford a GTX1070, better yet GTX1080 for a more consistent FPS at max settings.

To be fair there is this one AOC monitor that's only $332.77 (1080p 144Hz).

But generally yeah I agree. Also depends on how much you're bothered by screen tearing.
post #459 of 735
Quote:
Originally Posted by epic1337 View Post

"G-Sync" with a GTX1060? you serious? the monitor is way more expensive than the card, its not even funny that you could afford a decent 144Hz monitor for half the price.

if you could afford a G-Sync monitor, you'd obviously can afford a GTX1070, better yet GTX1080 for a more consistent FPS at max settings.

The funny part being someones money would be better spent on a Gsync gaming monitor then a higher end gpu. Gsync makes your gpu hit above its weight.
     
CPUCPUMotherboardGraphics
i7-7820x H115i TUF X299 Mark 1 GTX 1080 ti 
RAMHard DriveHard DriveHard Drive
G.SKILL TridentZ RGB Series 32GB @ 3600 Samsung 960 evo 500gb m.2 Samsung 850 evo 500GB 1Tb HDD 
OSMonitorKeyboardPower
Win 10 pro Acer Predator X34 (34" 21:9 Gsync) DAS Keyboard 4 Professional SeaSonic X Series X-850 (SS-850KM3 Active PFC F3)  
CaseMouseAudioAudio
Fractal Design Define R5 Blackout with Window Logitech G700s ONKYO TX-NR525 (Receiver) Polk Audio CS2 Series II (Center) 
AudioAudio
Polk Audio Monitor40 Series II (x4) Polk Audio PSW Series PSW505 (12" Sub) 
CPUMotherboardGraphicsRAM
i5 4790k MSI Z97A Gaming 6 EVGA 1050 Ti Corsair Vegence Pro 16gb (2400) 
Hard DriveHard DriveCoolingOS
Kingston V300 (OS) Seagate 1tb hybrid hdd Cooler Master Hyper 212 Evo Windows 10 
MonitorPowerCase
XB270HU (1440p/144Hz/Gsync) RM850 Rosewill R5 
  hide details  
     
CPUCPUMotherboardGraphics
i7-7820x H115i TUF X299 Mark 1 GTX 1080 ti 
RAMHard DriveHard DriveHard Drive
G.SKILL TridentZ RGB Series 32GB @ 3600 Samsung 960 evo 500gb m.2 Samsung 850 evo 500GB 1Tb HDD 
OSMonitorKeyboardPower
Win 10 pro Acer Predator X34 (34" 21:9 Gsync) DAS Keyboard 4 Professional SeaSonic X Series X-850 (SS-850KM3 Active PFC F3)  
CaseMouseAudioAudio
Fractal Design Define R5 Blackout with Window Logitech G700s ONKYO TX-NR525 (Receiver) Polk Audio CS2 Series II (Center) 
AudioAudio
Polk Audio Monitor40 Series II (x4) Polk Audio PSW Series PSW505 (12" Sub) 
CPUMotherboardGraphicsRAM
i5 4790k MSI Z97A Gaming 6 EVGA 1050 Ti Corsair Vegence Pro 16gb (2400) 
Hard DriveHard DriveCoolingOS
Kingston V300 (OS) Seagate 1tb hybrid hdd Cooler Master Hyper 212 Evo Windows 10 
MonitorPowerCase
XB270HU (1440p/144Hz/Gsync) RM850 Rosewill R5 
  hide details  
post #460 of 735
Quote:
Originally Posted by GorillaSceptre View Post

You might want to post some evidence of your "predictions" and mahigans failed ones before calling him out, or anyone else for that matter.. Polaris performs pretty close to where he said it would iirc.. It's the perf/watt that caught a lot of people by surprise.

Everything is speculation, no one knows exactly what is happening besides AMD/Nvidia. Guessing a few things is meaningless as even a broken clock is right twice a day..

Nvidia having an advantage everywhere except DX12 is a bit of an oxymoron wouldn't you say? A large part of Nvidias advantage in perf/Watt is because they cut out most of the stuff/don't have what makes AMD strong using DX12 in the first place..

How do you know what margins AMD is getting on P10? We also don't know what happened with the 14nm process they used, it's pretty clear something went wrong considering how much people can under-volt P10.. Vega will be using TSMC as far as i know, one chip in the Polaris family on a different process to Nvidia doesn't give us any insight into the future/ show us where Polaris would of stacked up against Pascal if they were both using the same process.

Some of the quotes.

http://www.overclock.net/t/1591625/guru3d-nvidia-slide-reveals-numbers-on-single-and-double-precision-for-flagship-pascal-gpu/40#post_24906129

"And I ball parked Polaris as well... Worst case is around 16-17 Tflops (without taking into consideration the higher fp32 performance per CU of Polaris over Fiji).

Polaris will beat Pascal. Not because I want it too but because it will."


http://www.overclock.net/t/1592431/anand-ashes-of-the-singularity-revisited-a-beta-look-at-directx-12-asynchronous-shading/580#post_24953038

"Baffin/XT is Polaris 11. Greenland/XT is Vega 11.

BaffintXT is likely only 232mm2 which is about equivalent to a 464mm2 GPU on 28nm. So in terms of performance per watt, it is 2.5x that of an equivalent AMD Fiji GPU on the 28nm process assuming similar clockspeeds.

AMD have stated that Polaris' performance gain are 70% due to the 14LPP process and 30% due to architectural improvements. So it's not a big GPU but it packs a punch. An AMD Hawaii part (290/290x) or Grenada part (390/390x) has a die size of 438mm2. So that should give you an idea of the relative density of Polaris 11.

So Polaris 11 will likely be faster than Fiji (Fury/FuryX) but will likely also include less Stream Processors (thus less CUs)."

He recently try to back track and realign these to performance at 390x levels and fury x level. And did a little Fib. Warning: Spoiler! (Click to show)
"False...

I was talking about Vega and not Polaris. I speculated that Polaris 10 would output around FuryX performance. The RX 480 appears to be around 390x performance which means that the 480x will probably hit the Fury X performance figures."

He also said Polaris would be getting HBM.

http://www.overclock.net/t/1594965/oc3d-amd-polaris-10-engineering-sample-pictured/70#post_24999000

He also predicted that Polaris would have better performance per watt than pascal.

http://www.overclock.net/t/1597918/tweaktown-amds-next-gen-polaris-gpus-to-be-used-in-apple-desktops-and-notebooks/80#post_25101841 Warning: Spoiler! (Click to show)
"I'm laughing biggrin.gif

Polaris has the capacity to power gate at the CU-level. Meaning that unused ALUs switch off, used ALUs boost (run at a high frequency).

CUs also adapt to work loads on the fly. Meaning, not all SIMDs will be used or organized the same way. It will entirely depend on the work loads.

In other words, Polaris will be quite revolutionary in terms of perf/watt. Yes, better than Pascal in this respect."

Considering the rx 480 is the full chip, he went from expecting better than fury x performance, better than pascal in certain scenarios. To eventually down to 390x performance for the cutdown variant of polaris 10 when it was becoming clear his prediction were off big time, even his lowered expectations as things became clear. Compared to reality, he was quite off. As mentioned in his post this was because he predicted this massive IPC gain of 40%. E.g 5tflops of polaris is equal to 7 tflops of fiji. Add in the performance per watt claims and he was off big time. He fed the hype machine.

Compare this to my prediction and solid logic which produced a more accurate prediction.

People need to pull back their expectations big time. A 232mm2 14nm GPU with Hawaii like performance(a 440mm2 die) is much more realistic, Particularly since all the 390x versions have been clocked very high because of the maturity of the chip and has that huge 512bit bus aiding it.

http://www.overclock.net/t/1594675/pcper-amds-raja-koduri-talks-moving-past-crossfire-smaller-gpu-dies-hbm2-and-more/70#post_24992080
Warning: Spoiler! (Click to show)
It would be very difficult to make a die that's 25% faster than a fury x with a die only 232mm2. You have to remember performance over a certain point is going to be limited by a 256bit bus. Getting over gtx 980 ti performance and Fury x, I think is impossible with a 256bit bus. Particularly because AMD color compression isn't as good as Nvidia. Average performance of the 380x is lower than the 280x according to techpowerup, not by much but it is lower.

In addition, it's a 232mm2 die. That's already a giant limiting factor on performance.

AMD was only able to beat 398mm2 made on 40nm with a 212mm2 made on 28nm die by 10% in performance. This was with a completely new architecture to boot. On top of this, Cayman or the 398mm2 die still had dp functionality.

The Fury x is a 600mm2 die, which has HBM to save die space because of the more basic memory controller(without it, it would be a 650-700mm2 die(it's Tonga x 2 which is a 360mm2 die)), is running water cooling to add to the performane and sacrifaced double precision for gaming performance. Polaris 10 a 232mm2 die, beating fury x with a ddr5 card restricted to a 256 bit bus by 25-30% is a fantasy. Particularly when you add in this isn't a new architecture, it's an evolution of a existing one.

Add in Fire pro duo would be obsolete considering a polaris x2, would be cheaper to manufacture, equipable with 8gb of memory.

People need to pull back their expectations big time. A 232mm2 14nm GPU with Hawaii like performance(a 440mm2 die) is much more realistic, Particularly since all the 390x versions have been clocked very high because of the maturity of the chip and has that huge 512bit bus aiding it.


People need to remember people don't upgrade every 2 years like overclock.net. There is alot of people on keplar and pitcairns that upgrade every or 5 years and polaris 10 is for those people.

Here I predicted that AMD was exaggerating their improvements and the possibility of AMD exceeding their 120watt goals

What this ultimately translates into is 390-390x levels of performance for 140-150 watts. This is a realistic expectation given marketing exaggeration and what information they have given us and taking into account deficits of last generation. What AMD fans fail to realize generally is that Tonga is so far behind gm204 as far as performance per watt, that reaching gtx 980 ti levels of performance in 120-140 watt chip would need AMD to basically triple their performance per watt and a real one, not a fake marketing derived one.

http://www.overclock.net/t/1598515/game-debate-rumour-amd-polaris-10-reportedly-offers-near-980-ti-performance-for-300-usd/480#post_25139694

I also predicted samsungs 14nm node wasn't going to have superior performance characteristics for desktop and high performance use vs TSMC.

http://www.overclock.net/t/1587480/wccftech-nvidia-s-pascal-is-mia-could-be-in-trouble-reports-allege-drive-px2-demoed-with-gtx-980m-instead-of-pascal/10#post_24786826
Warning: Spoiler! (Click to show)
Samsung doesn't have the experience of TSMC on high performance processes. It all about low power, low performance and frequency which is built for mobile. In fact global foundaries even has more experience than samsung in this regard, and GF has consistently crapped the bed for the most part.

Samsung getting the high performance node right the first time would be a real surprise considering their lack of experience and there nodes and experience, specialized for mobile application rather than desktop and laptop application.

I also predicted that AMD wasn't going to launch a limited lineup of GPU on finfet for the year.

http://www.overclock.net/t/1584297/tomshardware-amds-next-gen-products-taped-out-possibly-delayed/40#post_24712015

If you read my posts I stressed pitcairns class chip with hawaii like performance. I also predicted the largest variant of polaris was the 232mm2 when people were still fantasizing about something tahiti sized.

http://www.overclock.net/t/1594256/digitimes-amd-expected-to-unveil-polaris-gpu-in-june/20#post_24980545

I predicted the limited IPC gains which contradicted mahigans 40% gain ones.

What I did get wrong is I thought AMD was going to keep fury around alot longer, because AMD is too poor of a company to amortize the Fiji product so quickly. However this seems to be a result of pascal.

I didn't expect Nvidia to value price their 1070 with it's level of performance and torpedo the 600mm2 cards and raise the amount of performance for the dollar so much. Although it doesn't exist at 379, for marketing purposes it does and if Nvidia wished, they could have partners release cards at that price. I am also pretty sure, AMD didn't and it's why the price of polaris is so low. Since the 1070 is 50 percent faster than it, it has to cost atleast 50 percent more for Polaris to keep it's value proposition.
Edited by tajoh111 - 7/6/16 at 11:48am
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