GTX 1060 leak (TAKE WITH A TRUCKLOAD OF SALT)

[PlayStation 2]

I see it barely edging out the RX 480, if it does.

Probably gonna be a really close fight like the 960 vs. 285/380.

[Briggsy]

2 minutes ago, Dan Castellaneta said:

I see it barely edging out the RX 480, if it does.

Probably gonna be a really close fight like the 960 vs. 285/380.

1060 will certainly be more efficient, as the 1080 draws the same power as the 480. There's really no getting around that.

[PlayStation 2]

12 minutes ago, Briggsy said:

1060 will certainly be more efficient, as the 1080 draws the same power as the 480. There's really no getting around that.

I was comparing much more to its actual performance.

Efficiency wise, oh yeah, the 1060 will whip the 480's ass.

[Deli]

1 hour ago, Briggsy said:

1060 will certainly be more efficient, as the 1080 draws the same power as the 480. There's really no getting around that.

But they are both low TDP cards. Why would people pay some much attention to the efficiency? It isn't like the 20w difference will suddenly heats up your room, or adds load to the power bill, or you will need a new beefier PSU to power it.

[SSL]

13 minutes ago, Deli said:

But they are both low TDP cards. Why would people pay some much attention to the efficiency? It isn't like the 20w difference will suddenly heats up your room, or adds load to the power bill, or you will need a new beefier PSU to power it.

 

Why not? Certainly it will make a difference for small form-factor.

[Deli]

1 hour ago, SSL said:

 

Why not? Certainly it will make a difference for small form-factor.

People put 980Ti in SFX build, just saying.

[SSL]

1 minute ago, Deli said:

People put 980Ti in SFX build, just saying.

 

No doubt.

[Briggsy]

53 minutes ago, Dan Castellaneta said:

I was comparing much more to its actual performance.

Efficiency wise, oh yeah, the 1060 will whip the 480's ass.

I suppose it is possible to extrapolate some kind of number by comparing a 1070 to a 980, and making a wild guess from that to get the performance of a 1060.

 

**Take this post with 2 truckloads of salt**

 

for starters:

GTX 1070 = 1920 cores

GTX 980 = 2048 cores

 

------------------

 

980 is ~71% slower than 1070 so:

71% / 2048

= 0.034667% performance per core

 

GTX 1070:

100% / 1920

= 0.05208% performance per core

 

So if the math holds a gtx 960 should be ~37% of a 1070's performance.

(0.034667 * 1024)

 

meh, close enough.

Spoiler

 

 

------------

So therefore, a GTX 1060 (if it indeed has 1280 cuda cores) should be:   **drumroll**

1280 * 0.05208

= 66.66% of a 1070.

 

The RX 480 is 66.22% of a 1070 (100% / 151%)

 

Conclusion: the 1060 will probably beat a 480 by about Margin of error percent, but it could be either way by a couple percent if the rumor of higher boost clocks on the 1060 is true. The upcoming 480 power nerf might also cause the 1060 to edge ahead slightly, but not by as much as Nvidia would like to claim.

[zMeul]

14 hours ago, Vode said:

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.

the Async Compute on GeForces is done at the hardware level, the scheduling it's not .. it's done preemptively

[Humbug]

3 hours ago, Briggsy said:

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.

Polaris is as efficient as 28nm Maxwell.

https://www.techpowerup.com/reviews/AMD/RX_480/25.html

It's a lot less efficient than Pascal.

[zMeul]

6 minutes ago, Humbug said:

Polaris is as efficient as 28nm Maxwell.

https://www.techpowerup.com/reviews/AMD/RX_480/25.html

It's a lot less efficient than Pascal.

 

taking into account that Polaris 10 is on 14nm and Maxwell v2 is on 28nm .... Polaris should be way more efficient than that - it's a 2x node shrink and yet, they can't get that power usage right

[Megazero]

3 minutes ago, zMeul said:

To convert it into a more readable format:

Power draw per 100000000 transistors: (by order)

GTX 1070: 2.24W

GTX 1080: 2.55W

RX 480    : 2.91W

GTX 970  : 2.96W

R9 390     : 3.71W

Basically the 480 is a bit more power efficient than the 970, less efficient compare to Pascal and alot more compare to a R9 390

[zMeul]

2 minutes ago, Megazero said:

To convert it into a more readable format:

Power draw per 100000000 transistors: (by order)

GTX 1070: 2.24W

GTX 1080: 2.55W

RX 480    : 2.91W

GTX 970  : 2.96W

R9 390     : 3.71W

Basically the 480 is a bit more power efficient than the 970, less efficient compare to Pascal and alot more compare to a R9 390

[leadeater]

3 hours ago, SSL said:

 

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

Price is the biggest indicator of performance, has been and forever will be. When I saw the prices come out for the RX 480 before the reviews and it was the same as the cheapest 970's my first though was "Well those are going to be about the same performance", that is just how economics works.

 

A company will price a product for the best possible return while keeping the consumer demand high enough. Price it too high and it won't sell and price it too low you will not make enough per unit, and it also gives an inherent low quality feel to the product also further reducing sales.

 

There is a ton of psychology around product sales and right now Nvidia is nailing it, has been for a while.

[Marinatall_Ironside]

1 hour ago, leadeater said:

Price is the biggest indicator of performance, has been and forever will be. When I saw the prices come out for the RX 480 before the reviews and it was the same as the cheapest 970's my first though was "Well those are going to be about the same performance", that is just how economics works.

 

A company will price a product for the best possible return while keeping the consumer demand high enough. Price it too high and it won't sell and price it too low you will not make enough per unit, and it also gives an inherent low quality feel to the product also further reducing sales.

 

There is a ton of psychology around product sales and right now Nvidia is nailing it, has been for a while.

Agreed. Anybody who has taken an economics class in high school should know this.

[BirdyTheMighty]

and it will only cost just about $600

[Vode]

3 hours ago, zMeul said:

the Async Compute on GeForces is done at the hardware level, the scheduling it's not .. it's done preemptively

The Scheduling is the Asyncronous part of async compute and the compute is of course done on the hardware level. What's your point?

[Humbug]

3 hours ago, zMeul said:

taking into account that Polaris 10 is on 14nm and Maxwell v2 is on 28nm .... Polaris should be way more efficient than that

Yes you're right it should. It's not a big deal on desktop but it will really hold back AMD on the laptop parts.

 

Unless polaris 11 and the smaller parts are more efficient (which remains a possibility since Polaris 10 desktop was a latter design that stretched the architecture to it's limits). Time will tell.

[AlTech]

5 hours ago, zMeul said:

 

taking into account that Polaris 10 is on 14nm and Maxwell v2 is on 28nm .... Polaris should be way more efficient than that - it's a 2x node shrink and yet, they can't get that power usage right

I think it has something to do with the fact that it's GloFo Chips. They're not using 14nm SOC silicon here. They're using 14nm High Performance and High power Silicon.

[LAwLz]

6 hours ago, leadeater said:

Price is the biggest indicator of performance, has been and forever will be. When I saw the prices come out for the RX 480 before the reviews and it was the same as the cheapest 970's my first though was "Well those are going to be about the same performance", that is just how economics works.

 

A company will price a product for the best possible return while keeping the consumer demand high enough. Price it too high and it won't sell and price it too low you will not make enough per unit, and it also gives an inherent low quality feel to the product also further reducing sales.

 

There is a ton of psychology around product sales and right now Nvidia is nailing it, has been for a while.

Ehm, no? 

If that was the case then price:peformamce would never increase. But luckily for us we used to get big increases in price:performance ever generations up until we hit the 28nm wall. Mid range cards for lower prices than the previous generation high end cards were often trading blows in terms of performance, and with lots of other benefits too. 

 

If I hadn't been a huge pessimist when it comes to AMD promises (blame their CPU division for that) then I would totally have expected 390X performance across the board (not just if you cherry pick some titles). That would have been based on the improvements we saw in previous generations.

New architecture + die shrink (supposedly two generations but it's more like 1 to 1.5) should have equaled huge improvements. 

 

The improvements we got in places like efficiency, AMD was light-years behind in and are just now catching up to Nvidia's previous generation stuff. 

 

If you are me, the RX should have been better all around. Even at the 240 dollar price point. 

[Ansau]

Wowow, how much bias in the Polaris vs Maxwell...

First of all, you should be doing the comparison with 980 vs 480, both full chips, and not with the 970 which is a cut version.

Then, most reviews show the power consumption of the rx 480 is around the same, if not a bit lower than the 970, which means those 166W of the 480, but 154W of the 970 aren't fair.

 

If we put the power consumption of the 480 at 150W:

Polaris: 

- 2.78W per 10.000.000 transistors.

- 36.000.000 transistors/watt.

Maxwell:

- 3.17W per 10.000.000 transistors.

- 31.515.151 transistors/watt.

 

Polaris is about 15% more power efficient than Maxwell. And that's with Maxwell at reference frequencies (1127/1216 Mhz).

And Pascal is about 10% more power efficient than Polaris (Again comparing full chips 1080 vs 480).

[leadeater]

3 hours ago, LAwLz said:

Ehm, no? 

If that was the case then price:peformamce would never increase. But luckily for us we used to get big increases in price:performance ever generations up until we hit the 28nm wall. Mid range cards for lower prices than the previous generation high end cards were often trading blows in terms of performance, and with lots of other benefits too. 

 

If I hadn't been a huge pessimist when it comes to AMD promises (blame their CPU division for that) then I would totally have expected 390X performance across the board (not just if you cherry pick some titles). That would have been based on the improvements we saw in previous generations.

New architecture + die shrink (supposedly two generations but it's more like 1 to 1.5) should have equaled huge improvements. 

 

The improvements we got in places like efficiency, AMD was light-years behind in and are just now catching up to Nvidia's previous generation stuff. 

 

If you are me, the RX should have been better all around. Even at the 240 dollar price point. 

If the RX 480 had performed as well as some of the rumors then after launch existing pricing would have been adjusted, which has not happened. Price indicates performance not dictates. How long has the existing price structure been around? It's no coincidence that the buy cost of each performance segment has remained relatively static.

 

What was rather unique about the RX 480 was it's not the top end product, which usually sets the high watermark for performance then every product below it slots in to each pricing tier and as so price:performance goes up. The flaw here is you have to wait for the yard stick to measure against.

 

The general point was that the RX 480 could not have been priced more than the 970, AMD knew that. Would you buy the RX 480 if it cost more than another product but performed less? If the RX 480 was significantly faster than the 970 then it would have been priced higher as people would have paid that higher amount.

 

When it comes to the crunch the marketing and finance divisions of the company do a ton of cost break analysis and market trend assessment which they use to set the MSRP, their interest is solely on the best return. We can throw around all the technology and scientific improvements we like but in reality that has little to do with pricing, it should but it doesn't.

 

As a long time AMD/ATI user, 290X/6970/4850/X800GTO AGP/9600, I was watching with keen interest and hoping that it did very well but I don't let that cloud reasoned thought.

[Briggsy]

2 hours ago, Ansau said:

If we put the power consumption of the 480 at 150W:

These are average consumption numbers taken from the reviewers that actually do extensive power testing across a lot of games.

 

guru3D - 164W 480 vs. 154W 970

Techpowerup - 163W 480 vs. 156W 970

Kitguru (system load) - 225W 480 vs. 235W (209W adjusted) Palit 970 (which pulls 26 watts more than reference 970 in gaming)

 

I looked through a dozen other reviews, but most either do not list the 970 in use, only show peak power consumption (a somewhat useless metric here because the reference 970/980 have 2x 6-pin connectors, AIB's 8+6 pin), or their power consumption testing was done using a single synthetic bench or game instead of an average taken from a list of games - so they are not useful for anything but cherry-picking.

 

People compare the 480 to the 970 because that's where the performance matches, but really if you want to compare to a 980, thats even worse for the 480 because reference 970 and 980 draw the same average power while gaming, making the 480 efficiency even worse.

[zMeul]

7 hours ago, Vode said:

The Scheduling is the Asyncronous part of async compute and the compute is of course done on the hardware level. What's your point?

you make it sound like GeForces don't do the workload at the hardware level - they do

[Briggsy]

17 minutes ago, zMeul said:

you make it sound like GeForces don't do the workload at the hardware level - they do

Yeah, there is definitely hardware involvement with pascal in the scheduling. I think some people are so hung up on the idea of Pascal being a shrunk Maxwell, they fail to see the improvements in architecture.

 

Quote

http://www.pcper.com/reviews/Graphics-Cards/GeForce-GTX-1080-8GB-Founders-Edition-Review-GP104-Brings-Pascal-Gamers/Async

 

Pascal is the first GPU architecture to implement a pixel level preemption capability for graphics. The graphics units will keep track of their intermediate progress on the current rendering workload so that they can stop, save their state and move off the hardware to allow for the preempted workload to be addressed quickly. NVIDIA tells us the entire process of context switching can occur in less than 100 microseconds after the last pixel shading work is finished.