GTX 1060 leak (TAKE WITH A TRUCKLOAD OF SALT)

[DXMember]

12 minutes ago, LAwLz said:

So you do not think the 1060 will be 40% more efficient than the RX 480? I wouldn't be surprised if it is even higher than that, since the other cards with the same architecture and manufacturing process are much higher than that. 

 

We will see. 

I'm not saying it's not, or that it is

I'm saying it's a guessing game

[LAwLz]

4 minutes ago, DXMember said:

I'm not saying it's not, or that it is

I'm saying it's a guessing game

Yes, a guessing game where we already have two out of the 3 biggest components of what determines performance:watt. You might think the two biggest ones are "performance" and "watt", but they are actually architecture, manufacturing process and clock rate.

As we saw with the 1070 to 1080, the efficiency does not go up or down that much when you scale everything up/down. Assuming the GTX 1060 is just scaled down, not scaled down with a much higher frequency, we should expect similar performance:watt as the other cards in the family, which are right now about 70% more efficient than the RX 480.

 

Of course it is a "guessing game" since the results aren't in yet, but we can make guesses based on facts. If all our evidence points towards one thing, then it is a pretty safe guess to make.

 

I think you are trying to be objective here, but it's kind of backfiring. It's like when a Christian says evolution is just a "guess", and therefore their guess that we were all created should be treated as equal.

 

But luckily for us we aren't talking about creationism vs evolution here, so we will get our answer in like 2 weeks (hopefully). Until then, wanna join us and guess what efficiency will be like? There is no harm in guessing.

[DXMember]

45 minutes ago, LAwLz said:

You might think the two biggest ones are "performance" and "watt", but they are actually architecture, manufacturing process and clock rate.

clock rate doesn't scale linearly to power consumption - Both R9 Nano and Radeon Pro Duo demonstrate that perfectly

die size is a just as big of a factor as manufacturing process as shown with Hawaii and Tonga

[LAwLz]

10 minutes ago, DXMember said:

clock rate doesn't scale linearly to power consumption - Both R9 Nano and Radeon Pro Duo demonstrate that perfectly

die size is a just as big of a factor as manufacturing process as shown with Hawaii and Tonga

Yes I know it doesn't scale linearly with power consumption. That's why I said it was the third major aspect that determines performance:watt.

 

Die size is a terrible indicator for performance:watts.

The RX 480 is pretty similar to a 6870 in terms of die size (the 6870 is 10% larger). Are you really going to tell me that there should only be a 10% performance:watt ratio difference between the RX 480 and the HD 6870? Wanna know why their performance:watt numbers are so drastically difference despite having a very similar die size? Because architecture and manufacturing process are WAAAAAAAAAAAY more important for performance:watt than die size.

Remember, we are not talking about performance, nor are we talking about power consumption. We are talking about efficiency.

[leadeater]

On 7/2/2016 at 8:55 AM, LAwLz said:

Notice any pattern and abnormality?

We usually go 1 USD = 10 SEK and that accounts for VAT, import taxes and all other expenses that affects the price. The RX 480 costs about 25% more than it should,

Same thing here in NZ, the RX 480 and GTX 970 are exactly the same price down to the cent.

[Senzelian]

14 hours ago, Ryan_Vickers said:

Where they don't put one of the axis at 0.  It allows them to greatly exaggerate minor differences between different products, which can be useful for seeing those differences if, for some reason, you want to distinguish between 16100 and 16300, but when stacking cards up against each other for the purpose of showing that card B is X times faster than card A, this is NOT the right type of chart to use, since obviously, what people want to get out of that is a comparison of relative magnitudes and not a detailed showing of the top tips of the bars.

@Deli was talking about the MSRP of the 1080, not the chart.

[Vode]

Efficiency is also about the GPU archetecture itself. AMD still has those hardware schedulers in there while Nvidia does not. It's a trade off because there are benefits for Async compute if done on a hardware level. On the flip side the additional complexity manifests itself in the power draw.

 

Two different philosophies. AMD is going for DX12, Vulkan optimized hardware async compute while Nvidia is going with software async compute and DX11 optimized hardware.

[Megazero]

13 minutes ago, Vode said:

Efficiency is also about the GPU archetecture itself. AMD still has those hardware schedulers in there while Nvidia does not. It's a trade off because there are benefits for Async compute if done on a hardware level. On the flip side the additional complexity manifests itself in the power draw.

 

Two different philosophies. AMD is going for DX12, Vulkan optimized hardware async compute while Nvidia is going with software async compute and DX11 optimized hardware.

What do you mean when you said Hardware architecture schedulers? You mean AMD still have Vega? If that the case then Nvidia also have Volta already scheduled. Vega will compete against Pascal's GTX 1070 1080, while Volta will probably compete against Navi as they both scheduled for 2018.

Or if you talk about the way they do async, then yeah, I agree. Nvidia goes for a more brute force method with software while AMD do hardware async.

 

[KillDog]

Well, I HOPE AMD don't use GloFo's manufacturing process again in Vega. The efficiency is seriously bad, though it might be because of Async Compute Engines too.

Anyways, it will always be the way it has been: NVidia aiming for the current generation and having great efficiency as well as great performance in current games and succeeds, even though their cards are over-priced(Not talking about 970 here), while AMD aims for the future(again) and does not have good efficiency all round, but it's performance keeps improving thanks to drivers, but eventually they fail to regain their part in the market and even though their cards are cheap, no one buys it.

 

I do hope what I say is not true as we need AMD to remain in the market, else it will be the end of new technological evolutions, and both Intel and Nvidia will keep over-pricing their products.

[Vode]

8 minutes ago, Megazero said:

What do you mean when you said Hardware architecture schedulers? You mean AMD still have Vega? If that the case then Nvidia also have Volta already scheduled. Vega will compete against Pascal's GTX 1070 1080, while Volta will probably compete against Navi as they both scheduled for 2018.

Or if you talk about the way they do async, then yeah, I agree. Nvidia goes for a more brute force method with software while AMD do hardware async.

 

Hatdware schedulers is about async, yes. And they draw additional wattage. Nvidia got rid of the with the 580. Fermi and the notoriously hot and inefficient 480 still had them.

 

Which doesn't say that Hardware based Async isn't ineffiecent per se, it just is ineffiecient because DX11 doesn't make use of it. For DX12, Mantle and Vulkan that isnn't the case.

[AlTech]

On ‎7‎/‎1‎/‎2016 at 0:28 PM, gilang01 said:

snip

I'm calling Bullshit. There's no way it performs that fast under any circumstance. DX11 or DX12 or Vulkan, there's no way it can perform that fast.

 

It has the same amount of cores as the 970.

 

And why do I get the feeling this will have a 128 Bit bus instead of 256 Bit? 

On ‎7‎/‎1‎/‎2016 at 0:48 PM, gilang01 said:

hahaha ok. 

Anyone who finished high school should at least be able ton understand a simple graph tho 

Anybody who has finished high school should have better spelling and less typos.

[byalexandr]

Yeah sorry but it's definitely fake. As evil as Nvidia is, they wouldn't straight up call the 480 shit to AMD's face. Marketing yo

[LAwLz]

2 minutes ago, AluminiumTech said:

I'm calling Bullshit. There's no way it performs that fast under any circumstance. DX11 or DX12 or Vulkan, there's no way it can perform that fast.

 

It has the same amount of cores as the 970.

And according to this "leak", it also runs at 70% higher frequency. 20% extra performance over the GTX 970, with 70% higher clocks and the same number of cores, with a slightly tweaked architecture does not sound unreasonable to me. In fact it actually seems to be on the low side.

 

for 70% higher clocks and a slightly different architecture does not sound unreasonable to me, and that's what the graphs are showing.

[MrDynamicMan]

On 01/07/2016 at 7:37 PM, DXMember said:

RIP in peace...

Rest in peace in peace? 

 

Seeeeeeeems Legit. 

[DXMember]

5 minutes ago, MrDynamicMan said:

Rest in peace in peace? 

 

Seeeeeeeems Legit. 

RIP in pepperony?

HDD disk?

RAM memory?

[gilang01]

760 was slower than the 280/285, the 960 was slower than the 380....

Will the 1060 follow that route?

 

[Humbug]

2 hours ago, KillDog said:

Well, I HOPE AMD don't use GloFo's manufacturing process again in Vega. The efficiency is seriously bad, though it might be because of Async Compute Engines too.

 

Vega will be a lot more efficient than Polaris and not just because of HBM 2.

 

According to Raja Koduri (lead architect) the limitations on Polaris are not because of GloFo. They are because Polaris was originally supposed to be only for laptops. i.e. Polaris 11. Only subsequently they decided to stretch it and make a desktop part i.e. Polaris 10.

 

[MrDynamicMan]

49 minutes ago, DXMember said:

RIP in pepperony?

HDD disk?

RAM memory?

Fite me in irl

Solid state SDD 

CPU processor. 

 

[ace_cheaply]

57 minutes ago, MrDynamicMan said:

Fite me in irl

Solid state SDD 

CPU processor. 

 

Come at me bro.

 

I'm at the Atm machine with my PIN number.

 

[agentagent47]

Fellow Indonesian here! True  the markup here is ridiculously high and stupid. nevertheless i will buy this GPU since it is rumoured that it will be out on 7th of july and ramadhan finished on 12th of july. the tech store will be open on 13th, so there's quite a lot of chance that they will already have 1060 when they open ( if the GPU actually release on 7th   ) 

 

am i an Nvidia Fanboy ? maybe....just maybe

[vanished]

6 hours ago, ace_cheaply said:

Come at me bro.

 

I'm at the Atm machine with my PIN number.

 

RPM Package Manager anyone?  

[sof006]

Take Nvidias graphs with a HUMONGOUS PILE OF SALT and a cup of acid to wash it down with...  

[SSL]

2 minutes ago, sof006 said:

Take Nvidias graphs with a HUMONGOUS PILE OF SALT and a cup of acid to wash it down with...  

 

Why? Takes no great stretch of the imagination to place a $250+ card at 15% ahead of the 480.

[Briggsy]

8 hours ago, Humbug said:

Vega will be a lot more efficient than Polaris and not just because of HBM 2.

 

According to Raja Koduri (lead architect) the limitations on Polaris are not because of GloFo. They are because Polaris was originally supposed to be only for laptops. i.e. Polaris 11. Only subsequently they decided to stretch it and make a desktop part i.e. Polaris 10.

Makes sense that polaris so far isn't the efficiency king we were led to believe, the 480 is less efficient than the 28nm Maxwell architecture. Although, polaris 11 based on tests they showed last year was greatly more efficient than a 950/960 (can't remember which card), so Maybe there's hope for Vega efficiency yet.

 

I think the thing that bugs me the most, is the thought that AMD will go for the 300W "balls out" card. If Vega is chewing through power and only matching a 1080, that's a giant fail at 14nm. Then again, Raja didn't design Polaris or Vega, so I only hold him to his word when it comes to Navi, which I am excited to see. Raja might be awkward as hell, but hes a genius. Then again, so are Nvidia engineers.

[AlwaysFSX]

27 minutes ago, SSL said:

Why? Takes no great stretch of the imagination to place a $250+ card at 15% ahead of the 480.

RX 480 will perform like a 980! I can believe that!

GTX 1060 will perform like a 980! No, that's not right, take that with a huge truckload of salt.

 

Ppl pls