News AMD Announces Radeon VII

[Hitman928]

More info to come. . .

 

[maddie]

They aren't. Power draw increases as heat increases. This has been shown in several tests by various sites.

Important to keep in mind for the other upcoming 7nm GPUs & CPUs. Is this a characteristic of the new node? A steeper Power/Temp curve?

 

[Feld]

Important to keep in mind for the other upcoming 7nm GPUs & CPUs. Is this a characteristic of the new node? A steeper Power/Temp curve?

Nope, it's always been the case on every node. The hotter a chip is, the more power is required to properly push electrons through it. It's a characteristic of physical reality.

 

[coercitiv]

It's the fans and the LCD lighting.

It's not the fans. One 120 mm Noctua fan @ 3000 RPM is rated for a maximum input power of 3.6W. The fans on Radeon VII are smaller than that and together consume well under 10W when running at 3000 RPM.

Important to keep in mind for the other upcoming 7nm GPUs & CPUs. Is this a characteristic of the new node? A steeper Power/Temp curve?

I doubt it's that steep, there's probably more to that 80W difference than just static leakage delta.

 

[guskline]

I stand corrected that the nearly 80W delta between a stock Rad VII and a Watercooled Rad VII is due solely to the lack of 3 fans and a LED.

I suspect the lesser amount of heat due to watercooling amounts for the bulk of the delta.

 

[Topweasel]

It was how you could tell the original Vega and now this are clocked within an inch of its life. The dramatic drops in power usage at lower clocks and corresponding lowered voltage really show how much past their efficiency curve these guys are clocked.

 

[Mopetar]

It was how you could tell the original Vega and now this are clocked within an inch of its life. The dramatic drops in power usage at lower clocks and corresponding lowered voltage really show how much past their efficiency curve these guys are clocked.

It's not quite that simple. Most Vega buyers were able to undervolt their cards and get even better clock speeds at the same time. The issue is that AMD picks voltage values to maximize the number of chips that can hit a particular clock speed, which gives them the most silicon to sell. The unfortunate side effect is that most of their cards are being fed more power than they realistically needs which makes them inefficient even beyond architectural considerations.

 

[Topweasel]

It's not quite that simple. Most Vega buyers were able to undervolt their cards and get even better clock speeds at the same time. The issue is that AMD picks voltage values to maximize the number of chips that can hit a particular clock speed, which gives them the most silicon to sell. The unfortunate side effect is that most of their cards are being fed more power than they realistically needs which makes them inefficient even beyond architectural considerations.

That was kind of my point. Yeah I get that AMD is overvolting on top of just being miles over the prime spot on the efficiency curve. Just saying (and not really well now that I re-read it) that people testing the Vega 10 cards saw dramatic drops in power usage when they lowered the power and clocks just a little bit, or just by adding better cooling. GN I think got almost a 570's worth of power savings by playing around with different cooling options.

 

[maddie]

Well let's hope that the CPU & APU teams share their power control tech with the graphics teams. Night & day difference between divisions.

 

[guskline]

I will be interested in the results of other watercooled Radeon VIIs to see if this drop in power is consistent.