AMD Polaris Rumors and Discussion

[Deleted member 13524]

Interesting that guy above couldn't get the card to a stable 1500Mhz core w/ a watercooling setup, he had to drop to 1480Mhz, power limited?

As a customer, I'm a lot more disappointed that MSRP didn't go down even a tiny bit than the fact that 1500MHz can't be reached on watercooling but stock aircooled models can apparently reach >1400MHz with ease (which AFAIR only the NITRO+ models could do).
Like the 290->390 transition, it looks like the audience is getting a lot more gang-for-buck by purchasing a discounted RX400 in early April than by getting an equivalent RX500 in late April. And this time there isn't even the 2*VRAM excuse across the board.


I would get these prices if the cards were sporting 9GB/s memory and the new chips were clocking e.g. at 1400MHz base with >1500MHz boost, but as it stands AMD is simply offering less value here.
This is not what new card releases should be about, IMO.

 

[ImSpartacus]

Videocardz is claiming that the 500 series is released.

https://videocardz.com/68586/amd-launches-radeon-rx-500-series

Looks like the refreshed Polaris 10 XT will be referred to as Polaris 20 XTX and Polaris 20 XTR, which appear to be identical except the XTR variant is heavily binned to get to 1.4+ GHz core clocks.

And we're apparently only looking at 185W TDPs (which explains the 8-pin power connector).

This is encouraging for Vega's clock situation.

 

[CSI PC]

Videocardz is claiming that the 500 series is released.

https://videocardz.com/68586/amd-launches-radeon-rx-500-series

Looks like the refreshed Polaris 10 XT will be referred to as Polaris 20 XTX and Polaris 20 XTR, which appear to be identical except the XTR variant is heavily binned to get to 1.4+ GHz core clocks.

And we're apparently only looking at 185W TDPs (which explains the 8-pin power connector).

This is encouraging for Vega's clock situation.

Need to compare reference to reference, 1266MHz boost of 480 to 1340 boost of 580.
So that is a 6% boost clock increase with 15% greater TBP; I am using the 160W figure as well for the 480 rather than 150W.
I agree though lets see how Tom's/hardware.fr/pcper do with the reviews and power measurements with clock frequency/voltage behaviour.
Cheers

 

[Razor1]

185 watts for stock 1340, that is crazy. That is similar to what we saw of the rx 480 at 1340 mhz

Yeah gtx 1080 power draw for a mid range card.

 

[no-X]

Need to compare reference to reference, 1266MHz boost of 480 to 1340 boost of 580.
So that is a 6% boost clock increase with 15% greater TBP; I am using the 160W figure as well for the 480 rather than 150W.
I agree though lets see how Tom's/hardware.fr/pcper do with the reviews and power measurements with clock frequency/voltage behaviour.
Cheers

The TDP is more tied to the base clock than boost clock. Base clock of RX 570 was increased by 248 MHz and base clock of RX 580 o by 137 MHz. Higher TDP may be needed to ensure, that the GPU won't clock bellow the base clock even during high load.

 

[Razor1]

The TDP is more tied to the base clock than boost clock. Base clock of RX 570 was increased by 248 MHz and base clock of RX 580 o by 137 MHz. Higher TDP may be needed to ensure, that the GPU won't clock bellow the base clock even during high load.

Yeah that is pretty much how AMD rates their TDP recently.

 

[CSI PC]

The TDP is more tied to the base clock than boost clock. Base clock of RX 570 was increased by 248 MHz and base clock of RX 580 o by 137 MHz. Higher TDP may be needed to ensure, that the GPU won't clock bellow the base clock even during high load.

That was in the past not now.
The TDP is related to average with an application not base these days, it is not the same as Hawaii due to the way the Boost behaviour now works.
The 150W was meant to be boost average not at base clocks, it was made more confusing because AMD broke their TDP to 165ishW.
You really think there was only 10-14W difference between base and boost for the 480?

That average TBP is also backed up by PCPer (if one appreciates the use of a pass filter), the data both provide is very accurate.
Cheers

 

[CSI PC]

The TDP is more tied to the base clock than boost clock. Base clock of RX 570 was increased by 248 MHz and base clock of RX 580 o by 137 MHz. Higher TDP may be needed to ensure, that the GPU won't clock bellow the base clock even during high load.

This chart emphasises my point:

The 'official' TDP was 150W with a base clock of 1120Mhz and boost of 1266MHz.
AMD revised their power-dynamic boost envelope after the testing was done by those 3 sites that can measure it accurately and reduced it if one wanted to abide by the official TDP and PCIe.
Also AMD were basing the average on an application that was more efficient with use of power (stressed it less) and also at the lower resolution of 1080p (less power demand needed), although this all only applied since Polaris and the change to dynamic Boost behaviour.

Cheers

 

[CSI PC]

The TDP is more tied to the base clock than boost clock. Base clock of RX 570 was increased by 248 MHz and base clock of RX 580 o by 137 MHz. Higher TDP may be needed to ensure, that the GPU won't clock bellow the base clock even during high load.

My response has been moved to reviews, although context was regarding this point lol and not an actual review
https://forum.beyond3d.com/threads/amd-rx580-reviews.60052/#post-1977965
https://forum.beyond3d.com/threads/amd-rx580-reviews.60052/#post-1977967

Point being it changed with Polaris and now is based upon the average Boost clock due to how that now works.
Cheers

 

[Deleted member 13524]

As predicted, RX480 being flashed to RX580:

https://www.techpowerup.com/forums/threads/xfx-rx480-8gb-flashed-to-saphire-rx580.232489/

 

[Razor1]

AMD press presentation I think.

Wow they are comparing the 570 to the 1050ti, aren't they in different price brackets entirely.....

 

[entity279]

As predicted, RX480 being flashed to RX580:

https://www.techpowerup.com/forums/threads/xfx-rx480-8gb-flashed-to-saphire-rx580.232489/

AMD themselves didn't claim that they added anything to the 580. So any result other than this would have been a major surprise

 

[Deleted member 13524]

AMD themselves didn't claim that they added anything to the 580. So any result other than this would have been a major surprise

According to anandtech, AMD is claiming a "new memory state", which basically means the chips now have the ability to run multi-monitor setups and decode high bitrate videos without having to clock and configure the voltage of the GDDR5 to the same levels as if it's running a 3D game.

 

[entity279]

Hm yeah, I didnt asume that was implemented in the gpu hardware. I'd nominate the driver for that

 

[xEx]

the 580 is just an OC 480 so...Looking at the power consumption I dont even think these cards are respin it appears that they got more yields and because of that they clock them higher. Would have been better just to call it 480x calling it 580 is misleading is the same chip with higher quality pcb but I doubt the PC of the "cheap" 580 are better than the expensive 480...I dont like this move from AMD they went full Nvidia in this one.

 

[no-X]

the 580 is just an OC 480 so...Looking at the power consumption I dont even think these cards are respin

It seems the process enhancement affects power consumption at lower clocks more than power consumption at full speed. 2D (idle) power consumption of RX 580 is (and according to AMD should be) lower compared to RX 480: https://www.computerbase.de/2017-04/radeon-rx-580-570-test/4/

 

[Rootax]

I still don't see the point of a refresh like that...

 

[entity279]

Sure, let's see call it a rebrand

 

[xEx]

Sure, let's see call it a rebrand

Yeah and i have no idea why a 480x / 470x would be enough even in the past AMD had added value to their refresh cards in this one the only difference beside a more mature process is the different box...

If they don't have resources to improve the cards ok it can be understood but don't try to fool you're customers with dirty tactic like this, its not right.

 

[AlNom]

I had the impression that rebrands are a consequence of OEMs wanting something new every year, and the slower pace of progress/increased costs to fabbing new designs is partially responsible on there.