AMD Moves Vega 56 Embargo Forward, Asks Reviewers to Prioritize Over 64

[meenmeen1103]

2 minutes ago, Goku-sama said:

The fact they believe that the Vega 56 is the only card thats gonna captivate the most audience is so disappointing.

Thumbed up for this comment, as I agree, but mostly because of awesome sig

[Agosto]

5 hours ago, Coaxialgamer said:

However, we do have the gtx 1050 series. 

Unlike the rest of pascal, gp107 was made using Samsung’s 14nm LPP tech, the same as polaris and zen. 

And we find that it pretty much reaches very similar clocks, when considering that it likely has a different voltage curve ( to stay within TDP , which is more important than on the 1060 and up) 

and has a very small die. 

But if you look at the max voltage, voltage curve and frequencies (thanks TPU) you'll see the difference. Even if it has lower boost clocks, 1050&Ti need more voltage at both idle and load states (1.1V for 1050, 1.06 for all other pascal cards), have higher idle clocks and lower boost ones. IIRC they're also locked to 1911 max boost clock via driver, even those with the additional 6 pin connector.

Yes, I'd say TSMC 16nm is a better process than GloFo 14nm for discrete graphics cards.

[Coaxialgamer]

45 minutes ago, Agost said:

But if you look at the max voltage, voltage curve and frequencies (thanks TPU) you'll see the difference. Even if it has lower boost clocks, 1050&Ti need more voltage at both idle and load states (1.1V for 1050, 1.06 for all other pascal cards), have higher idle clocks and lower boost ones. IIRC they're also locked to 1911 max boost clock via driver, even those with the additional 6 pin connector.

Yes, I'd say TSMC 16nm is a better process than GloFo 14nm for discrete graphics cards.

And there are plenty of explanations for that.  Being this is the only chip produced for the process, gp107 could easily have been an afterthought. Plus, nvidia locking the chip doesn't help it. 

 

Lower yields on this specific chip could mean nvidia was forced to increase voltage a bit. You'll note that it is quite a bit denser. Smaller chips are also the least efficient (look at 1060 vs 1080 as an example). 

 

I should mention that 14nm LPP is being used for IBM's POWER9. 

As stated, no high performance chips have been made on 16FF.

[Agosto]

1 hour ago, Coaxialgamer said:

I should mention that 14nm LPP is being used for IBM's POWER9. 

As stated, no high performance chips have been made on 16FF.

And that looks like it's going to have a max clock of... yeah, 4 GHz, just like Ryzen.

16nm FF+ is already used by high power chips like P100 and GP102.

[leadeater]

1 hour ago, Agost said:

And that looks like it's going to have a max clock of... yeah, 4 GHz, just like Ryzen.

16nm FF+ is already used by high power chips like P100 and GP102.

@Coaxialgamer

 

Another thing to be aware of is that Intel uses a larger node size for the actual CPU cores, I'm not sure how big it actually is but from looking in to this it seems all the smaller nodes have max clock barriers currently.

 

Quote

It is important to note that Intel does not manufacture a complete, true 14 nm processor, as only the base layer of Intel processors is manufactured at the latest process node, to improve their yields and which also results in "greater electron flow" or power consumption for their processors. For instance, a Quad-core + GPU GT2 Core i7 Skylake K processor contains 1.75 billion transistors on a 122 mm² die size, for a transistor density of approximately 14 million transistors per millimeter squared, while a Ryzen 7 1800X contains 4.8 billion transistors on a 192 mm² die size, for a transistor density of 25 million transistors per millimeter squared.[32][33][34][35]

https://en.wikipedia.org/wiki/14_nanometer

 

Edit:

For extra reference Haswell was 1.4 billion transistors.