Overclock.net › Forums › Industry News › Hardware News › [Various] AMD Radeon RX 570 and RX 580 Reviews
New Posts  All Forums:Forum Nav:

[Various] AMD Radeon RX 570 and RX 580 Reviews - Page 10

post #91 of 140
Quote:
Originally Posted by Clocknut View Post

This kind of rebrand is soo deceptive. Whats the point of rebranding if it falls back the same tier and same price?

Historically Rebranding usually come with dropping the SKU at least 1 tier and a price drop.

8800GTS was 2nd tier = 9800GTX (with GTX260,280 serving higher end)
GTX480 = GTX570
GTX 680 = GTX 770
7970 = 280x (290x being the top)
7870 = 270x

Had this polaris with rebranded into 570 & 560 respectively and comes with a price cut to $199 & 149(for 8GB version). Nobody will make a fuss about it.

GTX 570 was not a rebranded GTX 480.
GTX 570 had 320 bit mem. interface, 1280MB VRAM and 40 ROP vs 384bit, 1536MB VRAM and 48 ROP on GTX 480.

CHEERS..
     
CPUMotherboardGraphicsRAM
E6750 G0 @ 3.3GHz (1.25V LOAD) Asus P5KPL-VM G31 GeForce 8500GT @ 700/1600/445 Apacer 2x1GB 800MHz 
Hard DriveOptical DriveOSMonitor
Hitachi Deskstar 160GB (7200,8MB) LG DVD Win 7 Ultimate 64bit build 7601 LG 19" 
KeyboardPowerCaseMouse
5$ keyboard Cheap stuff Linkworld crappy/modded SmartBox cheap 
Mouse Pad
N/A 
  hide details  
Reply
     
CPUMotherboardGraphicsRAM
E6750 G0 @ 3.3GHz (1.25V LOAD) Asus P5KPL-VM G31 GeForce 8500GT @ 700/1600/445 Apacer 2x1GB 800MHz 
Hard DriveOptical DriveOSMonitor
Hitachi Deskstar 160GB (7200,8MB) LG DVD Win 7 Ultimate 64bit build 7601 LG 19" 
KeyboardPowerCaseMouse
5$ keyboard Cheap stuff Linkworld crappy/modded SmartBox cheap 
Mouse Pad
N/A 
  hide details  
Reply
post #92 of 140
Quote:
Originally Posted by Serandur View Post

It gets worse looking at the average power draw:





GCN needs to die already. It's needed to die for years at this point, since Maxwell first arrived. There's no excusing this nonsense, it's a woefully noncompetitive architecture at this point.

I don't agree at all.

Firstly, it's not GCN or the architecture that is the problem, it is the manufacturing process it is produced on that hampers it at these frequencies. Which is key here. Pushing Polaris to 1450Mhz is way, way past the point where it is efficient and operates optimally for performance per watt. You know that, I know that, and TPU know that.

Secondly, TPU's power readings are unusually high for Polaris and I voiced this concern around launch as well, where their readings were a lot higher than KitGuru's, for example. It comes down to which readings you trust. KG's had Polaris 10 drawing power that was closer to the GTX 1060.

Now we'll see how 'woefully non-competitive' GCN is with Vega. Here is an architecture for the high end that will be more optimal at higher frequencies around 1400-1500Mhz. If there is a similar performance per watt gap between it and equivalent Pascal cards then your claim would stand up.
post #93 of 140
Quote:
Originally Posted by Shatun-Bear View Post

Secondly, TPU's power readings are unusually high for Polaris and I voiced this concern around launch as well, where their readings were a lot higher than KitGuru's, for example. It comes down to which readings you trust. KG's had Polaris 10 drawing power that was closer to the GTX 1060.

TPU are likely one of the most reliable sources as they don't measure the full system power consumption in just one or two scenarios. Measuring from wall outlet, using a $15 tool, is inaccurate.
Edited by Smanci - 4/19/17 at 3:50am
 
SFF
(13 photos)
 
CPUMotherboardGraphicsRAM
i7 3770 Z77N-WIFI GTX 970 Turbo Kingston Valueram VLP 2x4GB  
Hard DriveHard DriveCoolingCooling
Kingston UV400 240GB ADATA XPG SX900 64GB Thermalright True Spirit 120i AC Accelero Mono Plus 
CoolingCoolingOSMonitor
Scythe Slip Stream 120mm PWM Scythe Slip Stream 120mm PWM Windows 8.1 Enterprise 64-bit Dell P2416D 
PowerCase
Cooler Master V450S Jonsbo U2 
  hide details  
Reply
 
SFF
(13 photos)
 
CPUMotherboardGraphicsRAM
i7 3770 Z77N-WIFI GTX 970 Turbo Kingston Valueram VLP 2x4GB  
Hard DriveHard DriveCoolingCooling
Kingston UV400 240GB ADATA XPG SX900 64GB Thermalright True Spirit 120i AC Accelero Mono Plus 
CoolingCoolingOSMonitor
Scythe Slip Stream 120mm PWM Scythe Slip Stream 120mm PWM Windows 8.1 Enterprise 64-bit Dell P2416D 
PowerCase
Cooler Master V450S Jonsbo U2 
  hide details  
Reply
post #94 of 140
Quote:
Originally Posted by Shatun-Bear View Post

I don't agree at all.

Firstly, it's not GCN or the architecture that is the problem, it is the manufacturing process it is produced on that hampers it at these frequencies. Which is key here. Pushing Polaris to 1450Mhz is way, way past the point where it is efficient and operates optimally for performance per watt. You know that, I know that, and TPU know that.

Secondly, TPU's power readings are unusually high for Polaris and I voiced this concern around launch as well, where their readings were a lot higher than KitGuru's, for example. It comes down to which readings you trust. KG's had Polaris 10 drawing power that was closer to the GTX 1060.

Now we'll see how 'woefully non-competitive' GCN is with Vega. Here is an architecture for the high end that will be more optimal at higher frequencies around 1400-1500Mhz. If there is a similar performance per watt gap between it and equivalent Pascal cards then your claim would stand up.

There is absolutely no good reason to believe the problem lies with the process node instead of the architecture. Polaris is more-or-less the same as the GCN3 microarchitecture we saw on TSMC's 28nm node, the same node on which Maxwell 2 (architecturally nearly identical to Pascal) similarly pummeled GCN in both power and die space efficiency. Push two competing chips of roughly comparable size on similar nodes from Maxwell/Pascal and GCN to equivalent points along respective maximum frequency scales and we've seen time and time again that Maxwell/Pascal, without fail, trounce GCN in efficiency. GCN is more-or-less dead in mobile for that very reason. Nvidia's own GP107 on an all but identical 14nm node to Polaris continues to demonstrate that the differences lie primarily in the microarchitectures and not, as some people repeatedly try to push, in the process nodes.

There are no ands, ifs, or buts about it. GCN as it is and has been is massively inferior in both power efficiency and die space maximization to Maxwell/Pascal. As for Vega, it will no doubt have the benefit of HBM power/die space savings as Fiji did before it, thereby skewing the efficiency differences to some degree even if it remains largely the same GCN we've already seen. Nevertheless, it has been repeatedly established that GCN is in fact a less efficient microarchitecture than Maxwell/Pascal. We have over three years of these products coinciding on the market at the same time repeatedly demonstrating it.

Example: Take even the RX 480 in its original form and it's more-or-less neck and neck with the stock 1060 in performance... but it uses 16% more die space and ~40% more power consumption to do so with no demonstrated reason as to why GloFo/Samsung's 14nm process should be to blame. You can easily find similar deficiencies in both metrics (performance/watt and performance/die space) on the 28nm node between Tonga and pretty much any Maxwell chip. That is the mark of an inherently less efficient microarchitecture.
Edited by Serandur - 4/19/17 at 4:44am
Ice-nine
(17 items)
 
  
CPUMotherboardGraphicsRAM
i7-6700K Asus Z170-DELUXE Gigabyte G1 GTX 980 Ti Corsair Vengeance LPX DDR4 3000 
Hard DriveHard DriveOptical DriveCooling
Samsung 850 Evo 500 GB Seagate Barracuda 7200 RPM 2 TB Asus DVD Drive DRW-24B1ST/BLK/B/AS Phanteks PH-TC14PE 
OSMonitorKeyboardPower
Windows 8.1 Pro AOC Q2770PQU Corsair K95 RGB - Cherry MX Blue EVGA Supernova G2 750W 
CaseMouseMouse PadAudio
NZXT Phantom 410 Gunmetal Edition Logitech G502 Proteus Core Razer Goliathus Speed Edition Small Cyber Acoustics Satellite CA-3602 
Audio
Oppo PM-3 Closed Back Planar Magnetic Headphones 
  hide details  
Reply
Ice-nine
(17 items)
 
  
CPUMotherboardGraphicsRAM
i7-6700K Asus Z170-DELUXE Gigabyte G1 GTX 980 Ti Corsair Vengeance LPX DDR4 3000 
Hard DriveHard DriveOptical DriveCooling
Samsung 850 Evo 500 GB Seagate Barracuda 7200 RPM 2 TB Asus DVD Drive DRW-24B1ST/BLK/B/AS Phanteks PH-TC14PE 
OSMonitorKeyboardPower
Windows 8.1 Pro AOC Q2770PQU Corsair K95 RGB - Cherry MX Blue EVGA Supernova G2 750W 
CaseMouseMouse PadAudio
NZXT Phantom 410 Gunmetal Edition Logitech G502 Proteus Core Razer Goliathus Speed Edition Small Cyber Acoustics Satellite CA-3602 
Audio
Oppo PM-3 Closed Back Planar Magnetic Headphones 
  hide details  
Reply
post #95 of 140
Quote:
Originally Posted by Serandur View Post

There is absolutely no good reason to believe the problem lies with the process node instead of the architecture. Polaris is more-or-less the same as the GCN3 microarchitecture we saw on TSMC's 28nm node, the same node on which Maxwell 2 (architecturally nearly identical to Pascal) similarly pummeled GCN in both power and die space efficiency. Push two competing chips of roughly comparable size on similar nodes from Maxwell/Pascal and GCN to equivalent points along respective maximum frequency scales and we've seen time and time again that Maxwell/Pascal, without fail, trounce GCN in efficiency. GCN is more-or-less dead in mobile for that very reason. Nvidia's own GP107 on an all but identical 14nm node to Polaris continues to demonstrate that the differences lie primarily in the microarchitectures and not, as some people repeatedly try to push, in the process nodes.

The differences are largely down to neither architecture or process nodes actually (I was only trying to point out Polaris at 1450Mhz, the highest boost clock from any of the partner 580s, shows Polaris' efficiency in a very bad light), but drivers and software, both areas where Nvidia, with vastly more cash, have a huge advantage.

Conversely, if you look at performance for Polaris cards vs Pascal/Maxwell in DX12 or Vulkan, the 'pummelling' is happening the other way around. This is proof that GCN is extremely competent actually:





So there are definitely ands, ifs' and buts.

Quote:
Example: Take even the RX 480 in its original form and it's more-or-less neck and neck with the stock 1060 in performance... but it uses 16% more die space and ~40% more power consumption to do so with no demonstrated reason as to why GloFo/Samsung's 14nm process should be to blame. You can easily find similar deficiencies in both metrics (performance/watt and performance/die space) on the 28nm node between Tonga and pretty much any Maxwell chip. That is the mark of an inherently less efficient microarchitecture.

Again, when software implementation gets in the way, Polaris doesn't look great, so the truth is same performance in DX11 but significantly faster in DX12/Vulkan. Also, 40% more power is way higher than what I've read, it again depends on what source you are using:

post #96 of 140
Quote:
Originally Posted by Shatun-Bear View Post

Warning: Spoiler! (Click to show)
The differences are largely down to neither architecture or process nodes actually (I was only trying to point out Polaris at 1450Mhz, the highest boost clock from any of the partner 580s, shows Polaris' efficiency in a very bad light), but drivers and software, both areas where Nvidia, with vastly more cash, have a huge advantage.

Conversely, if you look at performance for Polaris cards vs Pascal/Maxwell in DX12 or Vulkan, the 'pummelling' is happening the other way around. This is proof that GCN is extremely competent actually:





So there are definitely ands, ifs' and buts.
Again, when software implementation gets in the way, Polaris doesn't look great, so the truth is same performance in DX11 but significantly faster in DX12/Vulkan. Also, 40% more power is way higher than what I've read, it again depends on what source you are using:


You're cherrypicking two very specific and deliberately GCN-favoring scenarios where Polaris gets the slightest leg up and trying to pretend that it somehow reverses the efficiency gap here (it doesn't), let alone simply bridging it then (still doesn't; ~8% more performance for ~40% more power consumption in that 480 vs 1060 case), let alone in general. It doesn't even in those GCN-favoring examples. That 480, even in those best-case scenarios, is barely edging out the 1060 at all at disproportionately greater die/transistor/power cost. No pummeling going on there... not in the way you meant at least.

And no, it doesn't depend on what source you are using. That very image says "total system power consumption". That includes the substantial power draw of the CPU, which should be roughly the same in all those systems, and simply doesn't measure pure GPU power draw as Smanci mentioned TPU does above. That source does not in any way contradict TPU. Just as in TPU's testing, the 480's adding another ~50 watts over the 1060 and ending up consuming more than the 1070. Subtract ~40 watts from the end results to account for the difference between Kitguru's total system power consumption and TPU's GPU only power consumption and you have nearly the exact same results. This isn't really debatable. The numbers speak and they say GCN is an inferior microarchitecture to Maxwell/Pascal.
Edited by Serandur - 4/19/17 at 7:01am
Ice-nine
(17 items)
 
  
CPUMotherboardGraphicsRAM
i7-6700K Asus Z170-DELUXE Gigabyte G1 GTX 980 Ti Corsair Vengeance LPX DDR4 3000 
Hard DriveHard DriveOptical DriveCooling
Samsung 850 Evo 500 GB Seagate Barracuda 7200 RPM 2 TB Asus DVD Drive DRW-24B1ST/BLK/B/AS Phanteks PH-TC14PE 
OSMonitorKeyboardPower
Windows 8.1 Pro AOC Q2770PQU Corsair K95 RGB - Cherry MX Blue EVGA Supernova G2 750W 
CaseMouseMouse PadAudio
NZXT Phantom 410 Gunmetal Edition Logitech G502 Proteus Core Razer Goliathus Speed Edition Small Cyber Acoustics Satellite CA-3602 
Audio
Oppo PM-3 Closed Back Planar Magnetic Headphones 
  hide details  
Reply
Ice-nine
(17 items)
 
  
CPUMotherboardGraphicsRAM
i7-6700K Asus Z170-DELUXE Gigabyte G1 GTX 980 Ti Corsair Vengeance LPX DDR4 3000 
Hard DriveHard DriveOptical DriveCooling
Samsung 850 Evo 500 GB Seagate Barracuda 7200 RPM 2 TB Asus DVD Drive DRW-24B1ST/BLK/B/AS Phanteks PH-TC14PE 
OSMonitorKeyboardPower
Windows 8.1 Pro AOC Q2770PQU Corsair K95 RGB - Cherry MX Blue EVGA Supernova G2 750W 
CaseMouseMouse PadAudio
NZXT Phantom 410 Gunmetal Edition Logitech G502 Proteus Core Razer Goliathus Speed Edition Small Cyber Acoustics Satellite CA-3602 
Audio
Oppo PM-3 Closed Back Planar Magnetic Headphones 
  hide details  
Reply
post #97 of 140
Thread Starter 
Quote:
Originally Posted by Clocknut View Post

This kind of rebrand is soo deceptive. Whats the point of rebranding if it falls back the same tier and same price?

Historically Rebranding usually come with dropping the SKU at least 1 tier and a price drop.

8800GTS was 2nd tier = 9800GTX (with GTX260,280 serving higher end)
GTX480 = GTX570
GTX 680 = GTX 770
7970 = 280x (290x being the top)
7870 = 270x

Had this polaris with rebranded into 570 & 560 respectively and comes with a price cut to $199 & 149(for 8GB version). Nobody will make a fuss about it.
R9 390x was a R9 290x. and really if the product will be similar on the whole lineup, they dont need to reduce its category to entry level or low end because there are already GPUs which have that tag, also there isnt something faster to rename as a entry mid level
Quote:
Originally Posted by Serandur View Post

Warning: Spoiler! (Click to show)
And no, it doesn't depend on what source you are using. That very image says "total system power consumption". That includes the substantial power draw of the CPU, which should be roughly the same in all those systems, and simply doesn't measure pure GPU power draw as Smanci mentioned TPU does above. That source does not in any way contradict TPU. Just as in TPU's testing, the 480's adding another ~50 watts over the 1060 and ending up consuming more than the 1070. Subtract ~40 watts from the end results to account for the difference between Kitguru's total system power consumption and TPU's GPU only power consumption and you have nearly the exact same results. This isn't really debatable.
The numbers speak and they say GCN is an inferior microarchitecture to Maxwell/Pascal.
Only if you compare power consumption
Quote:
Originally Posted by tpi2007 View Post

I wasn't expecting this. 185w board power for the RX 580 and 150w for the RX 570, notwithstanding the respin, ends up being a brute force approach to gain more performance. I was expecting that the respin and better manufacturing process had allowed them to increase clockspeeds while maintaining the same power consumption or even a little less. In other words, I was expecting the same type of improvement that we saw on 32nm with the transition from Bulldozer to Piledriver.

Also, considering that they initially said (see below) back in the RX 400 series launch that there was a fix for the "high power consumption readings at idle and under partial loads", I hope that they make the new firmware that introduces an intermediary power state to improve media playback and multi-monitor power use available for RX 400 series owners, because as far as I know the fix mentioned below never arrived until now. And part of it is assuming that they actually improve multi-monitor power consumption, which many reviews show seems to not be working properly yet.
http://www.tomshardware.com/reviews/amd-radeon-rx-460,4707-5.html
Alreayd AMD was using Lower power FF14, i dont think that keeping the same design, aswell increasing voltage and clock speed it will reduce power draw
Edited by PontiacGTX - 4/19/17 at 7:20am
  
Reply
  
Reply
post #98 of 140
My System
(15 items)
 
  
CPUMotherboardGraphicsRAM
i7-6700K @ 4.6GHz ASUS Z170-A Gigabyte GTX 1070 G1 Gaming G.SKILL Ripjaws V Series DDR4 3000MHZ 16GB 
Hard DriveCoolingOSMonitor
Samsung EVO 850 250GB Noctua NH-D15S Windows 10 BenQ XL2720Z 
KeyboardPowerCaseMouse
Corsair K70 Corsair RM650x Corsair 600T Logitech G502 
  hide details  
Reply
My System
(15 items)
 
  
CPUMotherboardGraphicsRAM
i7-6700K @ 4.6GHz ASUS Z170-A Gigabyte GTX 1070 G1 Gaming G.SKILL Ripjaws V Series DDR4 3000MHZ 16GB 
Hard DriveCoolingOSMonitor
Samsung EVO 850 250GB Noctua NH-D15S Windows 10 BenQ XL2720Z 
KeyboardPowerCaseMouse
Corsair K70 Corsair RM650x Corsair 600T Logitech G502 
  hide details  
Reply
post #99 of 140
Quote:
Originally Posted by Omega X View Post

I think at this point telling people to wait a few weeks for Vega is useless at this point. They've already turned into electro-zombies hungry for new silicon.

True. Anyone else seen the rumor that vega will come with a watercooler like the fury x?

I think it has been so long now, no matter what amd release with vega, it will fail to impress.
post #100 of 140
Quote:
Originally Posted by Newbie2009 View Post

True. Anyone else seen the rumor that vega will come with a watercooler like the fury x?

I think it has been so long now, no matter what amd release with vega, it will fail to impress.

It will come with AIO?

biggrin.gifwheee.gif

Now I just need it to be faster than a 1080 and cost $399.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [Various] AMD Radeon RX 570 and RX 580 Reviews