Overclock.net › Forums › Industry News › Hardware News › [AdoredTV] Pascal vs Maxwell at same clockspeeds, same FLOPS (1080 vs 980 Ti)
New Posts  All Forums:Forum Nav:

[AdoredTV] Pascal vs Maxwell at same clockspeeds, same FLOPS (1080 vs 980 Ti) - Page 30

post #291 of 305
Quote:
Originally Posted by mtcn77 View Post

I'm starting to learn a few things in this thread after all. I forgot that "67,200 MPixel/s" meant '67.2 Gigapixels per second'. According to that, GCN is hitting 95-100% of the theoretically maximum fill rate capacity. Bandwidth utilization to the same extent, however, is nowhere to be seen. The theoretical utilisation rate between 290X(DC II Top spec) and Fury X is dropping from 76% to 65% respectively. It could relate to rising losses from increased magnitude, however wasn't HBM expected to avert that trend somewhat?
Clearly, each gpu shows higher utilisation at the step where the bottleneck is present. So far, I think all convictions that GCN is fillrate and geometry limited while Geforce is bandwidth limited are kind of vindicated looking at the available data(which is awesome by the way).
[Source]

HBM won't help GCN when the data hits a roadblock at the front end/back end. It's like having a 50 lane highway but you can only enter and exit from/to a gravel back road. Nvidia is like suburb driving at Mach 10 and always hitting traffic lights. You pretty much got the idea of what's going on though.

To further illustrate things , notice how neither Fiji nor Hawaii hit their bandwidth limit but Maxwell does. Like you said, you see higher use at the bottleneck


Edited by Echoa - 7/31/16 at 2:28am
Toxic-DT
(19 items)
 
Toxic-LT
(12 items)
 
Toxic-SV
(10 items)
 
CPUMotherboardGraphicsRAM
i7 4770k Gigabyte Z87X-UD4H Gigabyte GTX 1060 Gskill Ares 
Hard DriveHard DriveHard DriveOptical Drive
Mushkin Eco3 Kingdian  Corsair LE  Lite On 
CoolingOSMonitorKeyboard
Cryorig H5 Windows 10 pro Dell S2415h Azio MK Retro 
PowerCaseMouseAudio
Seasonic G Series 650w Define R5  Logitech G500s Asus Xonar DSX 
AudioAudioAudio
Monoprice Retro Heaphones Monoprice Vocal Dynamic Mic Edifier r980t  
CPUMotherboardGraphicsRAM
I7 920xm Asus g73jh Radeon HD 6990m 8gb Hynix DDR3 1333 
Hard DriveHard DriveCoolingOS
Seagate Laptop Thin 500gb Corsair Force LE  Stock G73jh Windows 10 
MonitorKeyboardPowerMouse
1600x900 Glossy G73jh Screen G73jh Backlit Keyboard Delta 150w PSU Verbatim Laptop mouse 
CPUMotherboardGraphicsRAM
Xeon e5649 HP z400 FirePro v4800 12gb Gkill Ripjaw 
Hard DriveHard DriveCoolingOS
Hitachi Ultrastar Western Digital 320gb HP High Performance Cooler Ubuntu Gnome 16.10 
PowerCase
HP z400 HP z400 
  hide details  
Reply
Toxic-DT
(19 items)
 
Toxic-LT
(12 items)
 
Toxic-SV
(10 items)
 
CPUMotherboardGraphicsRAM
i7 4770k Gigabyte Z87X-UD4H Gigabyte GTX 1060 Gskill Ares 
Hard DriveHard DriveHard DriveOptical Drive
Mushkin Eco3 Kingdian  Corsair LE  Lite On 
CoolingOSMonitorKeyboard
Cryorig H5 Windows 10 pro Dell S2415h Azio MK Retro 
PowerCaseMouseAudio
Seasonic G Series 650w Define R5  Logitech G500s Asus Xonar DSX 
AudioAudioAudio
Monoprice Retro Heaphones Monoprice Vocal Dynamic Mic Edifier r980t  
CPUMotherboardGraphicsRAM
I7 920xm Asus g73jh Radeon HD 6990m 8gb Hynix DDR3 1333 
Hard DriveHard DriveCoolingOS
Seagate Laptop Thin 500gb Corsair Force LE  Stock G73jh Windows 10 
MonitorKeyboardPowerMouse
1600x900 Glossy G73jh Screen G73jh Backlit Keyboard Delta 150w PSU Verbatim Laptop mouse 
CPUMotherboardGraphicsRAM
Xeon e5649 HP z400 FirePro v4800 12gb Gkill Ripjaw 
Hard DriveHard DriveCoolingOS
Hitachi Ultrastar Western Digital 320gb HP High Performance Cooler Ubuntu Gnome 16.10 
PowerCase
HP z400 HP z400 
  hide details  
Reply
post #292 of 305
Quote:
Originally Posted by Echoa View Post

They actually improved the ROP and front end with Polaris but we have yet to see how this will translate on the high end. Polaris manages to hit 500 shaders, 256bit bus, and 32rops above its weight class which is a good improvement so far.
not enough to make up a full 32rop loss. I highly doubt AMD double the ROP efficiency.

With the clock speed Polaris has, it should given a 48Rop. (Tahiti has decoup Rop from memory bus width, AMD could do it here again)
Gaming Rig
(10 items)
 
Work/Web Rig
(11 items)
 
Web Rig
(8 items)
 
CPUMotherboardGraphicsRAM
i5-2500K @4GHz stock volt OC Asus P8Z77-V LK Palit GTX 750Ti StormX Dual Corsair Vengence 1600MHz_CL8 4GBx4 
Hard DriveHard DriveCoolingOS
Seagate 80GB 7200rpm Maxtor 250GB 7200rpm Cooler Master Hyper 212+ Windows 10 Pro 
MonitorCase
BenQ XL2720z @ 1920x1080 144Hz Silverstone Ps06 
CPUMotherboardGraphicsRAM
Core 2 Quad Q9650 Gigabyte EP41-UD3L Asus 1GB Radeon 7790 DirectCU OC Corsair Gaming Ram 2GBx2 DDR2-800 
Hard DriveHard DriveCoolingOS
Samsung 80GB 7200rpm Western Digital 200GB 7200rpm Cooler Master Hyper TX3 Windows 7 pro 
MonitorPowerCase
Samsung 226bw 1680x1050 Acbel iPower 510w (450w PSU) LianLi PS05 
CPUMotherboardGraphicsRAM
AMD E350 MSI E350IA-E45 AMD Radeon 6310 4GB Kingston DDR3 1333 
Hard DriveOSMonitorPower
Western Digital 80GB Linux Slackware 14.1 Samsung 22" 1680x1050 LCD Acbel 300w PSU 
  hide details  
Reply
Gaming Rig
(10 items)
 
Work/Web Rig
(11 items)
 
Web Rig
(8 items)
 
CPUMotherboardGraphicsRAM
i5-2500K @4GHz stock volt OC Asus P8Z77-V LK Palit GTX 750Ti StormX Dual Corsair Vengence 1600MHz_CL8 4GBx4 
Hard DriveHard DriveCoolingOS
Seagate 80GB 7200rpm Maxtor 250GB 7200rpm Cooler Master Hyper 212+ Windows 10 Pro 
MonitorCase
BenQ XL2720z @ 1920x1080 144Hz Silverstone Ps06 
CPUMotherboardGraphicsRAM
Core 2 Quad Q9650 Gigabyte EP41-UD3L Asus 1GB Radeon 7790 DirectCU OC Corsair Gaming Ram 2GBx2 DDR2-800 
Hard DriveHard DriveCoolingOS
Samsung 80GB 7200rpm Western Digital 200GB 7200rpm Cooler Master Hyper TX3 Windows 7 pro 
MonitorPowerCase
Samsung 226bw 1680x1050 Acbel iPower 510w (450w PSU) LianLi PS05 
CPUMotherboardGraphicsRAM
AMD E350 MSI E350IA-E45 AMD Radeon 6310 4GB Kingston DDR3 1333 
Hard DriveOSMonitorPower
Western Digital 80GB Linux Slackware 14.1 Samsung 22" 1680x1050 LCD Acbel 300w PSU 
  hide details  
Reply
post #293 of 305
Quote:
Originally Posted by Clocknut View Post

not enough to make up a full 32rop loss. I highly doubt AMD double the ROP efficiency.

With the clock speed Polaris has, it should given a 48Rop. (Tahiti has decoup Rop from memory bus width, AMD could do it here again)

I never said they doubled it, but with 32 less ROP it is keeping pace with the 390/x tongue.gif it's fillrate is lower though, that's factual and it would like more ROPs. it hits a wall in certain scenarios and shows no improvement vs Tonga (texture filtering performs exactly like 380x). Polaris still shows improvement on Fiji's weak points, but it's a 200$card and we won't see the high end fruit till later
Toxic-DT
(19 items)
 
Toxic-LT
(12 items)
 
Toxic-SV
(10 items)
 
CPUMotherboardGraphicsRAM
i7 4770k Gigabyte Z87X-UD4H Gigabyte GTX 1060 Gskill Ares 
Hard DriveHard DriveHard DriveOptical Drive
Mushkin Eco3 Kingdian  Corsair LE  Lite On 
CoolingOSMonitorKeyboard
Cryorig H5 Windows 10 pro Dell S2415h Azio MK Retro 
PowerCaseMouseAudio
Seasonic G Series 650w Define R5  Logitech G500s Asus Xonar DSX 
AudioAudioAudio
Monoprice Retro Heaphones Monoprice Vocal Dynamic Mic Edifier r980t  
CPUMotherboardGraphicsRAM
I7 920xm Asus g73jh Radeon HD 6990m 8gb Hynix DDR3 1333 
Hard DriveHard DriveCoolingOS
Seagate Laptop Thin 500gb Corsair Force LE  Stock G73jh Windows 10 
MonitorKeyboardPowerMouse
1600x900 Glossy G73jh Screen G73jh Backlit Keyboard Delta 150w PSU Verbatim Laptop mouse 
CPUMotherboardGraphicsRAM
Xeon e5649 HP z400 FirePro v4800 12gb Gkill Ripjaw 
Hard DriveHard DriveCoolingOS
Hitachi Ultrastar Western Digital 320gb HP High Performance Cooler Ubuntu Gnome 16.10 
PowerCase
HP z400 HP z400 
  hide details  
Reply
Toxic-DT
(19 items)
 
Toxic-LT
(12 items)
 
Toxic-SV
(10 items)
 
CPUMotherboardGraphicsRAM
i7 4770k Gigabyte Z87X-UD4H Gigabyte GTX 1060 Gskill Ares 
Hard DriveHard DriveHard DriveOptical Drive
Mushkin Eco3 Kingdian  Corsair LE  Lite On 
CoolingOSMonitorKeyboard
Cryorig H5 Windows 10 pro Dell S2415h Azio MK Retro 
PowerCaseMouseAudio
Seasonic G Series 650w Define R5  Logitech G500s Asus Xonar DSX 
AudioAudioAudio
Monoprice Retro Heaphones Monoprice Vocal Dynamic Mic Edifier r980t  
CPUMotherboardGraphicsRAM
I7 920xm Asus g73jh Radeon HD 6990m 8gb Hynix DDR3 1333 
Hard DriveHard DriveCoolingOS
Seagate Laptop Thin 500gb Corsair Force LE  Stock G73jh Windows 10 
MonitorKeyboardPowerMouse
1600x900 Glossy G73jh Screen G73jh Backlit Keyboard Delta 150w PSU Verbatim Laptop mouse 
CPUMotherboardGraphicsRAM
Xeon e5649 HP z400 FirePro v4800 12gb Gkill Ripjaw 
Hard DriveHard DriveCoolingOS
Hitachi Ultrastar Western Digital 320gb HP High Performance Cooler Ubuntu Gnome 16.10 
PowerCase
HP z400 HP z400 
  hide details  
Reply
post #294 of 305
Quote:
Originally Posted by Echoa View Post

HBM won't help GCN when the data hits a roadblock at the front end/back end. It's like having a 50 lane highway but you can only enter and exit from/to a gravel back road. Nvidia is like suburb driving at Mach 10 and always hitting traffic lights. You pretty much got the idea of what's going on though.

To further illustrate things , notice how neither Fiji nor Hawaii hit their bandwidth limit but Maxwell does. Like you said, you see higher use at the bottleneck



Disagree.

HBM2 will still be very useful in reducing power consumption. On Polaris, we can see that the chip uses 110W, while the RAM chips use 30W. A 512 bit bus with GDDR5X would use a lot more power. HBM2 can dramatically reduce that, while increasing bandwidth.


Also, in the picture that you quoted:




We can see that:
  • Hawaii has no color compression and has a real bandwidth of 263 Gb/s (versus 320 Gb/s in theory) - 82.19% efficiency
  • The Fury X is actually very inefficient in this regard - a real bandwidth of 333 GB/s (versus a 512 Gb/s theoretical). There is some color compression going on, which gives it 387 Gb/s - very inefficient actually at 65.04%
  • The Titan X has 238 GBb/s of real bandwidth, on a 336 GB/s of theroetical. Not the best efficiency either at 70.83% actual.
  • Nvidia has considerably better color compression on Maxwell, so with a perfectly compressible image it becomes from 238 Gb/s to 379 Gb/s, almost matching the Fury X!


If you were to take the Pascal GTX 1080 and consider that it has even better Delta Color Compression:


Not sure why they got better results on Nvidia with the black texture this time compared with the Fury X review - perhaps new drivers or something else improved Nvidia's results.

We can see though that AMD has improved Delta Color Compression somewhat with Polaris.


Judging by the results though, it's not as good as Maxwell's, much less Pascal.

I suspect that Nvidia may be closing in on what is possible with Maxwell. We will see with Volta. I expect to see some continued improvements, but not as big as before in Delta Color Compression. Compared to Maxwell versus Kepler, Pascal is already a smaller improvement over Maxwell.

The bottom line though is that for real effective bandwidth, AMD needs HBM2. They also need to ensure that their HBM2 controller uses more than 65% of the theoretical bandwidth - ideally getting it over 80% like in Hawaii. The certainly need a much better front end this time around. Remember though, the main advantage of HBM2 is not the raw bandwidth, it is the power consumption and will allow more other parts on the die.

I suppose they could also work on Delta Color Compression. They will need to improve it for their mainstream cards, until HBM2 (or whatever new technology succeeds it, as they are planning to implement "NextGen memory" in Navi) becomes affordable that GPUs like the Rx480's successor and APUs can use HBM2.


Quote:
Originally Posted by Clocknut View Post

not enough to make up a full 32rop loss. I highly doubt AMD double the ROP efficiency.

With the clock speed Polaris has, it should given a 48Rop. (Tahiti has decoup Rop from memory bus width, AMD could do it here again)



I"d argue for as many as 64 ROPs (so 4 RBEs per Shader Engine). I just hope that the 4096 core part comes with more than 64 ROPs and 256 Z/Stencil ROPs this time.
Edited by CrazyElf - 7/31/16 at 8:37am
Trooper Typhoon
(20 items)
 
  
CPUMotherboardGraphicsGraphics
5960X X99A Godlike MSI 1080 Ti Lightning MSI 1080 Ti Lightning 
RAMHard DriveHard DriveHard Drive
G.Skill Trident Z 32 Gb Samsung 850 Pro Samsung SM843T 960 GB Western Digital Caviar Black 2Tb 
Hard DriveOptical DriveCoolingCooling
Samsung SV843 960 GB LG WH14NS40 Cryorig R1 Ultimate 9x Gentle Typhoon 1850 rpm on case 
OSMonitorKeyboardPower
Windows 7 Pro x64 LG 27UD68 Ducky Legend with Vortex PBT Doubleshot Backlit... EVGA 1300W G2 
CaseMouseAudioOther
Cooler Master Storm Trooper Logitech G502 Proteus Asus Xonar Essence STX Lamptron Fanatic Fan Controller  
  hide details  
Reply
Trooper Typhoon
(20 items)
 
  
CPUMotherboardGraphicsGraphics
5960X X99A Godlike MSI 1080 Ti Lightning MSI 1080 Ti Lightning 
RAMHard DriveHard DriveHard Drive
G.Skill Trident Z 32 Gb Samsung 850 Pro Samsung SM843T 960 GB Western Digital Caviar Black 2Tb 
Hard DriveOptical DriveCoolingCooling
Samsung SV843 960 GB LG WH14NS40 Cryorig R1 Ultimate 9x Gentle Typhoon 1850 rpm on case 
OSMonitorKeyboardPower
Windows 7 Pro x64 LG 27UD68 Ducky Legend with Vortex PBT Doubleshot Backlit... EVGA 1300W G2 
CaseMouseAudioOther
Cooler Master Storm Trooper Logitech G502 Proteus Asus Xonar Essence STX Lamptron Fanatic Fan Controller  
  hide details  
Reply
post #295 of 305
Quote:
Originally Posted by sugarhell View Post

Mods should lock this thread.

People watch the video first !

Second the use of the word fanboy is everywhere. Especially amd fanboys when this thread is only about nvidia

Third ocn sucks. We cant have a proper thread about a educational video that is really usefull for some people. Stop fighting and call each other fanboys when you cant even read a simple white paper available to everyone instead you believe the PR . Yeah lock this thread

Agree thumb.gif


Just allow posts after 7pm - kids will be in bed then biggrin.gif :

its in the title ' Pascal vs Maxwell at same clockspeeds,' but have learnt more about amd cards lachen.gif
Edited by ladcrooks - 7/31/16 at 9:42am
post #296 of 305
Quote:
Originally Posted by magnek View Post

When you guys say "bandwidth", do you mean "memory bandwidth"? Because Pascal (or GP104 at least) barely responds to memory bandwidth improvements:



http://www.anandtech.com/show/10325/the-nvidia-geforce-gtx-1080-and-1070-founders-edition-review/31

You'd probably see different results with a 1070 though - I think the GDDR5X helps the 1080 a lot in that department. 3DMark, at least, shows pretty substantial gains from overclocking 1070 memory.

Edit: Well, not as much as I thought, but still more than the 1080. About 3.5% gain from a 12.5% memory clock increase.
Edited by Forceman - 7/31/16 at 10:06am
post #297 of 305
Quote:
Originally Posted by CrazyElf View Post

Disagree.

HBM2 will still be very useful in reducing power consumption. On Polaris, we can see that the chip uses 110W, while the RAM chips use 30W. A 512 bit bus with GDDR5X would use a lot more power. HBM2 can dramatically reduce that, while increasing bandwidth.


Also, in the picture that you quoted:




We can see that:
  • Hawaii has no color compression and has a real bandwidth of 263 Gb/s (versus 320 Gb/s in theory) - 82.19% efficiency
  • The Fury X is actually very inefficient in this regard - a real bandwidth of 333 GB/s (versus a 512 Gb/s theoretical). There is some color compression going on, which gives it 387 Gb/s - very inefficient actually at 65.04%
  • The Titan X has 238 GBb/s of real bandwidth, on a 336 GB/s of theroetical. Not the best efficiency either at 70.83% actual.
  • Nvidia has considerably better color compression on Maxwell, so with a perfectly compressible image it becomes from 238 Gb/s to 379 Gb/s, almost matching the Fury X!


If you were to take the Pascal GTX 1080 and consider that it has even better Delta Color Compression:


Not sure why they got better results on Nvidia with the black texture this time compared with the Fury X review - perhaps new drivers or something else improved Nvidia's results.

We can see though that AMD has improved Delta Color Compression somewhat with Polaris.


Judging by the results though, it's not as good as Maxwell's, much less Pascal.

I suspect that Nvidia may be closing in on what is possible with Maxwell. We will see with Volta. I expect to see some continued improvements, but not as big as before in Delta Color Compression. Compared to Maxwell versus Kepler, Pascal is already a smaller improvement over Maxwell.

The bottom line though is that for real effective bandwidth, AMD needs HBM2. They also need to ensure that their HBM2 controller uses more than 65% of the theoretical bandwidth - ideally getting it over 80% like in Hawaii. The certainly need a much better front end this time around. Remember though, the main advantage of HBM2 is not the raw bandwidth, it is the power consumption and will allow more other parts on the die.

I suppose they could also work on Delta Color Compression. They will need to improve it for their mainstream cards, until HBM2 (or whatever new technology succeeds it, as they are planning to implement "NextGen memory" in Navi) becomes affordable that GPUs like the Rx480's successor and APUs can use HBM2.
I"d argue for as many as 64 ROPs (so 4 RBEs per Shader Engine). I just hope that the 4096 core part comes with more than 64 ROPs and 256 Z/Stencil ROPs this time.

I didn't say HBM/2 is completely useless, just that in regards to what AMDs actual issue are HBM isn't the remedy. In regards to power consumption and reducing board complexity HBM is great and by no means useless I by no means think it is.

even in best case scenario AMD can't reach its theoretical bandwidth vs pascal/maxwell where a nvidia does easily. GCN needs a better front end, better compression, and more/better ROPs which we seem to agree on. AMD doesn't NEED HBM (for bandwidth, power consumption yes though) , Polaris shows it can keep up with lower bandwidth vs it's higher bandwidth brothers. GCN needs the means to utilize that bandwidth it has, which it can't due to its weak front and back end. I too hope Vega has more as Fiji just couldn't adequately drive 4k shaders with its resources, that should also improve their bandwidth utilization. Even with the same shader could Vega should be another to improve on Fiji by a sizeable amount of they fix Fiji's problems

As for the topic, Pascal is very obviously a die shrink and clock boost of Maxwell, it's not bad but nothing to write home about either and we won't see real changes till Volta but we knew that already. Nvidia would've been risking more going with a brand new arch on a brand new node, even AMD went and tweaked/improved GCN 1.2 vs intro an entirely new arch on a new node. Can't be mad or surprised by either company doing it as is the smart move, just nvidia is also charging a kings ransom for it.
Edited by Echoa - 7/31/16 at 11:45am
Toxic-DT
(19 items)
 
Toxic-LT
(12 items)
 
Toxic-SV
(10 items)
 
CPUMotherboardGraphicsRAM
i7 4770k Gigabyte Z87X-UD4H Gigabyte GTX 1060 Gskill Ares 
Hard DriveHard DriveHard DriveOptical Drive
Mushkin Eco3 Kingdian  Corsair LE  Lite On 
CoolingOSMonitorKeyboard
Cryorig H5 Windows 10 pro Dell S2415h Azio MK Retro 
PowerCaseMouseAudio
Seasonic G Series 650w Define R5  Logitech G500s Asus Xonar DSX 
AudioAudioAudio
Monoprice Retro Heaphones Monoprice Vocal Dynamic Mic Edifier r980t  
CPUMotherboardGraphicsRAM
I7 920xm Asus g73jh Radeon HD 6990m 8gb Hynix DDR3 1333 
Hard DriveHard DriveCoolingOS
Seagate Laptop Thin 500gb Corsair Force LE  Stock G73jh Windows 10 
MonitorKeyboardPowerMouse
1600x900 Glossy G73jh Screen G73jh Backlit Keyboard Delta 150w PSU Verbatim Laptop mouse 
CPUMotherboardGraphicsRAM
Xeon e5649 HP z400 FirePro v4800 12gb Gkill Ripjaw 
Hard DriveHard DriveCoolingOS
Hitachi Ultrastar Western Digital 320gb HP High Performance Cooler Ubuntu Gnome 16.10 
PowerCase
HP z400 HP z400 
  hide details  
Reply
Toxic-DT
(19 items)
 
Toxic-LT
(12 items)
 
Toxic-SV
(10 items)
 
CPUMotherboardGraphicsRAM
i7 4770k Gigabyte Z87X-UD4H Gigabyte GTX 1060 Gskill Ares 
Hard DriveHard DriveHard DriveOptical Drive
Mushkin Eco3 Kingdian  Corsair LE  Lite On 
CoolingOSMonitorKeyboard
Cryorig H5 Windows 10 pro Dell S2415h Azio MK Retro 
PowerCaseMouseAudio
Seasonic G Series 650w Define R5  Logitech G500s Asus Xonar DSX 
AudioAudioAudio
Monoprice Retro Heaphones Monoprice Vocal Dynamic Mic Edifier r980t  
CPUMotherboardGraphicsRAM
I7 920xm Asus g73jh Radeon HD 6990m 8gb Hynix DDR3 1333 
Hard DriveHard DriveCoolingOS
Seagate Laptop Thin 500gb Corsair Force LE  Stock G73jh Windows 10 
MonitorKeyboardPowerMouse
1600x900 Glossy G73jh Screen G73jh Backlit Keyboard Delta 150w PSU Verbatim Laptop mouse 
CPUMotherboardGraphicsRAM
Xeon e5649 HP z400 FirePro v4800 12gb Gkill Ripjaw 
Hard DriveHard DriveCoolingOS
Hitachi Ultrastar Western Digital 320gb HP High Performance Cooler Ubuntu Gnome 16.10 
PowerCase
HP z400 HP z400 
  hide details  
Reply
post #298 of 305
Quote:
Originally Posted by TheLAWNOOB View Post

Its easy being AMD fanboy.

I think it's quite the opposite. AMD fans have taken a beating for years - then again maybe you're onto something: they could be so used to getting abused that it's now the norm for them..!
post #299 of 305
Quote:
Originally Posted by Echoa View Post

I didn't say HBM/2 is completely useless, just that in regards to what AMDs actual issue are HBM isn't the remedy. In regards to power consumption and reducing board complexity HBM is great and by no means useless I by no means think it is.

even in best case scenario AMD can't reach its theoretical bandwidth vs pascal/maxwell where a nvidia does easily. GCN needs a better front end, better compression, and more/better ROPs which we seem to agree on. AMD doesn't NEED HBM (for bandwidth, power consumption yes though) , Polaris shows it can keep up with lower bandwidth vs it's higher bandwidth brothers. GCN needs the means to utilize that bandwidth it has, which it can't due to its weak front and back end. I too hope Vega has more as Fiji just couldn't adequately drive 4k shaders with its resources, that should also improve their bandwidth utilization. Even with the same shader could Vega should be another to improve on Fiji by a sizeable amount of they fix Fiji's problems

As for the topic, Pascal is very obviously a die shrink and clock boost of Maxwell, it's not bad but nothing to write home about either and we won't see real changes till Volta but we knew that already. Nvidia would've been risking more going with a brand new arch on a brand new node, even AMD went and tweaked/improved GCN 1.2 vs intro an entirely new arch on a new node. Can't be mad or surprised by either company doing it as is the smart move, just nvidia is also charging a kings ransom for it.
What exactly is weak about its front end and back end..?
post #300 of 305
Quote:
Originally Posted by Audio View Post

haters gona hate. Reason this card sold so well is because it's a solid 30% faster then 980ti. If you can't get one gota justify it somehow with some research and youtube vids. Don't forget to cherry pick the benchmarks either.

I don't think he's hating on it, he just pointed out that nVidia were able to build a really efficient architecture in Maxwell which has allowed them to clock it faster in conjunction with a die shrink and product the goods that way...
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [AdoredTV] Pascal vs Maxwell at same clockspeeds, same FLOPS (1080 vs 980 Ti)