Overclock.net › Forums › Industry News › Hardware News › [pcper] NVIDIA Talks DX12, DX11 Efficiency Improvements
New Posts  All Forums:Forum Nav:

[pcper] NVIDIA Talks DX12, DX11 Efficiency Improvements - Page 13

post #121 of 203
Quote:
Originally Posted by Kuivamaa View Post

Check your own test numbers.
280X DX11 avg unit count 3961
280X Mantle avg unit count 4436
Titan DX11 avg unit count 4057

It most certainly isn't a directly comparable scenario (280X under mantle had more stuff going on on average). Even with the same amount of units I doubt we could possibly have identical twin runs. Mantle is obviously nondeterministic (AI does whatever it sees fit between different runs).

It does illustrate the point that nVidia drivers have less CPU overhead than AMD. The Titan, even though it had more units performed better than 280x in DX11. It also shows the larger, unavoidable DX11 overhead.
Millenium Falcon
(24 items)
 
  
CPUMotherboardGraphicsRAM
Core i7 4930k MSI Big Bang Xpower II EVGA GTX 690 Patriot Viper II Sector 7 
Hard DriveHard DriveHard DriveHard Drive
OCZ Deneva 2 Corsair Force 3 Maxtor Western Digital Green 
Optical DriveCoolingCoolingCooling
Samsung BD/DVD-RW Swiftech MCP655 x2 Black Ice GTX 480 Black Ice GTX 280 
CoolingCoolingCoolingCooling
Alphacool Repack Dual D5 Watercool Heatkiller 3.0 Alphacool GTX 690 fullcover Bitspower Big Bang Xpower II fullcover 
OSMonitorKeyboardPower
Windows 8.1 64-bit Professional 3x Dell S2340 Max Keyboard Durandal CoolerMaster V1000 
CaseMouseMouse PadAudio
Azza Genesis 9000B Logitech G700 Roccat Alumic Onkyo HT-S9100THX 
  hide details  
Reply
Millenium Falcon
(24 items)
 
  
CPUMotherboardGraphicsRAM
Core i7 4930k MSI Big Bang Xpower II EVGA GTX 690 Patriot Viper II Sector 7 
Hard DriveHard DriveHard DriveHard Drive
OCZ Deneva 2 Corsair Force 3 Maxtor Western Digital Green 
Optical DriveCoolingCoolingCooling
Samsung BD/DVD-RW Swiftech MCP655 x2 Black Ice GTX 480 Black Ice GTX 280 
CoolingCoolingCoolingCooling
Alphacool Repack Dual D5 Watercool Heatkiller 3.0 Alphacool GTX 690 fullcover Bitspower Big Bang Xpower II fullcover 
OSMonitorKeyboardPower
Windows 8.1 64-bit Professional 3x Dell S2340 Max Keyboard Durandal CoolerMaster V1000 
CaseMouseMouse PadAudio
Azza Genesis 9000B Logitech G700 Roccat Alumic Onkyo HT-S9100THX 
  hide details  
Reply
post #122 of 203
Quote:
Originally Posted by Kuivamaa View Post

Check your own test numbers.
280X DX11 avg unit count 3961
280X Mantle avg unit count 4436
Titan DX11 avg unit count 4057

It most certainly isn't a directly comparable scenario (280X under mantle had more stuff going on on average). Even with the same amount of units I doubt we could possibly have identical twin runs. Mantle is obviously nondeterministic (AI does whatever it sees fit between different runs).

The unit/etc. counts in the results text file mean very little if you have a custom scenario where it spawn the exact same number of units on the screen every time, from the same direction every time, and you see the fight happen from the same angle every time. If you're going to say it's an irrelevant test you should at least test it yourself first. Because you wont see any 10 fps differences. Or 5 fps differences, or 3 fps differences, probably not even 2 fps differences between runs.

I've run this thing a ton of times and with the same setup the results have always been within 1fps or so. 1.5fps being the biggest difference I think I saw.

It's from here: http://forums.overclockers.co.uk/showpost.php?p=25762826&postcount=94

And you have the same story there as well, the results are extremely consistent no matter what the txt file says afterwards about the details of the run. It's not just me who gets consistent fps results with these custom scenarios, other people do as well.

Custom scenarios are a different beast entirely. The fps results are consistent between runs (or as consistent as any canned bench you could have).

For example here's two runs I just did back to back:




So if even if you wanted to argue about that ~1fps difference you might get between runs (just like with any bench) it doesn't remove the huge difference between AMD and Nvidia DX11 CPU overhead.
Edited by Alatar - 3/24/14 at 6:00am
 
Benching
(17 items)
 
 
CPUMotherboardGraphicsRAM
[i7 5960X @ 4.8GHz] [Rampage V Extreme] [Titan 1400MHz (1500MHz bench)] [Various] 
Hard DriveCoolingCoolingCooling
[250GB 840EVO +2x SpinpointF3 1TB RAID0] [LD PC-V2 SS Phase Change] [XSPC X2O 750 pump/res] [Monsta 360 full copper + EK XT 360 + XT 240] 
MonitorPowerCaseAudio
[Crossover 27Q LED-P 1440p+ASUS 1200p+LG 1080p] [Corsair AX1200] [Dimastech Easy v3.0] [Sennheiser HD558s] 
CPUCPUMotherboardGraphics
FX 8320, FX 8350, Phenom II x2 555BE i7 3930K, i7 860, i7 4770K, 68x Celeron D CVF, commando, 2x RIVE, Z87X-OC Asus 4870x2, Sapphire 4870 
GraphicsGraphicsGraphicsGraphics
2x 5870, 5850, 5830, 5770 2x 3870x2, 3870 GTX Titan, GTX 480, GTX 590 GTX 285, GTX 260, 4x 9800GT, 8800GTX 
RAMHard DriveCoolingCooling
4x4GB vengeance, 2x4GB predatorX, 2x1GB OCZ DDR2 Intel X25-M 80GB LD PC-V2 SS Phase Change OCN Marksman 
CoolingCoolingOSPower
2x old tek slims (GPU) Various watercooling stuff win7, winxp AX1200 
Case
test bench / cardboard box 
  hide details  
Reply
 
Benching
(17 items)
 
 
CPUMotherboardGraphicsRAM
[i7 5960X @ 4.8GHz] [Rampage V Extreme] [Titan 1400MHz (1500MHz bench)] [Various] 
Hard DriveCoolingCoolingCooling
[250GB 840EVO +2x SpinpointF3 1TB RAID0] [LD PC-V2 SS Phase Change] [XSPC X2O 750 pump/res] [Monsta 360 full copper + EK XT 360 + XT 240] 
MonitorPowerCaseAudio
[Crossover 27Q LED-P 1440p+ASUS 1200p+LG 1080p] [Corsair AX1200] [Dimastech Easy v3.0] [Sennheiser HD558s] 
CPUCPUMotherboardGraphics
FX 8320, FX 8350, Phenom II x2 555BE i7 3930K, i7 860, i7 4770K, 68x Celeron D CVF, commando, 2x RIVE, Z87X-OC Asus 4870x2, Sapphire 4870 
GraphicsGraphicsGraphicsGraphics
2x 5870, 5850, 5830, 5770 2x 3870x2, 3870 GTX Titan, GTX 480, GTX 590 GTX 285, GTX 260, 4x 9800GT, 8800GTX 
RAMHard DriveCoolingCooling
4x4GB vengeance, 2x4GB predatorX, 2x1GB OCZ DDR2 Intel X25-M 80GB LD PC-V2 SS Phase Change OCN Marksman 
CoolingCoolingOSPower
2x old tek slims (GPU) Various watercooling stuff win7, winxp AX1200 
Case
test bench / cardboard box 
  hide details  
Reply
post #123 of 203
Quote:
Originally Posted by Alatar View Post

The unit/etc. counts in the results text file mean very little if you have a custom scenario where it spawn the exact same number of units on the screen every time, from the same direction every time, and you see the fight happen from the same angle every time. If you're going to say it's an irrelevant test you should at least test it yourself first. Because you wont see any 10 fps differences. Or 5 fps differences, or 3 fps differences, probably not even 2 fps differences between runs.

I've run this thing a ton of times and with the same setup the results have always been within 1fps or so. 1.5fps being the biggest difference I think I saw.

It's from here: http://forums.overclockers.co.uk/showpost.php?p=25762826&postcount=94

And you have the same story there as well, the results are extremely consistent no matter what the txt file says afterwards about the details of the run. It's not just me who gets consistent fps results with these custom scenarios, other people do as well.

Custom scenarios are a different beast entirely. The fps results are consistent between runs (or as consistent as any canned bench you could have).

For example here's two runs I just did back to back:




So if even if you wanted to argue about that ~1fps difference you might get between runs (just like with any bench) it doesn't remove the huge difference between AMD and Nvidia DX11 CPU overhead.

You are making a fair argument (as fair is to compare a 280X that trails badly a 290X in ROPs etc) on overheads but my point still stands. The bench you provided on the post above had similar unit count between both runs hence there similar fps. Not the case with your first post (10%+ variance there). If you can be bothered to run this several times till you get very similar unit count between runs with a radeon and a geforece we may have something more solid. Till then it is what it is, inconsistent.
Mastodon Ryzen
(12 items)
 
HP Z220
(8 items)
 
 
CPUMotherboardGraphicsRAM
R7 1800X Asus Crosshair VI Hero Sapphire RX Vega 64 reference Gskill TridentZ 
Hard DriveOptical DriveCoolingOS
Pny SSD 240GB Crucial MX100 CM Nepton 280L Win 10 
MonitorPowerCaseMouse
Acer Predator XG270HU Freesync XFX 750W Pro HAF XM Logitech G502 
CPUMotherboardGraphicsCooling
i7 3770 HP Quadro K2000 HP 
OSPowerCaseMouse
Win 7  HP 400W HP CMT RAT 7 
  hide details  
Reply
Mastodon Ryzen
(12 items)
 
HP Z220
(8 items)
 
 
CPUMotherboardGraphicsRAM
R7 1800X Asus Crosshair VI Hero Sapphire RX Vega 64 reference Gskill TridentZ 
Hard DriveOptical DriveCoolingOS
Pny SSD 240GB Crucial MX100 CM Nepton 280L Win 10 
MonitorPowerCaseMouse
Acer Predator XG270HU Freesync XFX 750W Pro HAF XM Logitech G502 
CPUMotherboardGraphicsCooling
i7 3770 HP Quadro K2000 HP 
OSPowerCaseMouse
Win 7  HP 400W HP CMT RAT 7 
  hide details  
Reply
post #124 of 203
I wont bother with installing the 280X again. As I told you, and showed you the variance between runs is very small no matter what the unit counts say. Not only my findings, other people have noted this as well. There's no reason to run it 10 times just to get the same results over and over again until the random avg. units numbers fit.

Even the mantle result is still very much representative. For obvious reasons listed above. The margin of error is not big enough to make the overall conclusion invalid in the slightest, even if the % might change by 1 or 2.

However since you don't want to believe that then so be it, but you can't deny that the DX11 results from both camps do show a very clear difference in DX11 CPU overhead. Hell, the Nvidia run has more average units, even though that makes next to no difference with this custom scenario.
 
Benching
(17 items)
 
 
CPUMotherboardGraphicsRAM
[i7 5960X @ 4.8GHz] [Rampage V Extreme] [Titan 1400MHz (1500MHz bench)] [Various] 
Hard DriveCoolingCoolingCooling
[250GB 840EVO +2x SpinpointF3 1TB RAID0] [LD PC-V2 SS Phase Change] [XSPC X2O 750 pump/res] [Monsta 360 full copper + EK XT 360 + XT 240] 
MonitorPowerCaseAudio
[Crossover 27Q LED-P 1440p+ASUS 1200p+LG 1080p] [Corsair AX1200] [Dimastech Easy v3.0] [Sennheiser HD558s] 
CPUCPUMotherboardGraphics
FX 8320, FX 8350, Phenom II x2 555BE i7 3930K, i7 860, i7 4770K, 68x Celeron D CVF, commando, 2x RIVE, Z87X-OC Asus 4870x2, Sapphire 4870 
GraphicsGraphicsGraphicsGraphics
2x 5870, 5850, 5830, 5770 2x 3870x2, 3870 GTX Titan, GTX 480, GTX 590 GTX 285, GTX 260, 4x 9800GT, 8800GTX 
RAMHard DriveCoolingCooling
4x4GB vengeance, 2x4GB predatorX, 2x1GB OCZ DDR2 Intel X25-M 80GB LD PC-V2 SS Phase Change OCN Marksman 
CoolingCoolingOSPower
2x old tek slims (GPU) Various watercooling stuff win7, winxp AX1200 
Case
test bench / cardboard box 
  hide details  
Reply
 
Benching
(17 items)
 
 
CPUMotherboardGraphicsRAM
[i7 5960X @ 4.8GHz] [Rampage V Extreme] [Titan 1400MHz (1500MHz bench)] [Various] 
Hard DriveCoolingCoolingCooling
[250GB 840EVO +2x SpinpointF3 1TB RAID0] [LD PC-V2 SS Phase Change] [XSPC X2O 750 pump/res] [Monsta 360 full copper + EK XT 360 + XT 240] 
MonitorPowerCaseAudio
[Crossover 27Q LED-P 1440p+ASUS 1200p+LG 1080p] [Corsair AX1200] [Dimastech Easy v3.0] [Sennheiser HD558s] 
CPUCPUMotherboardGraphics
FX 8320, FX 8350, Phenom II x2 555BE i7 3930K, i7 860, i7 4770K, 68x Celeron D CVF, commando, 2x RIVE, Z87X-OC Asus 4870x2, Sapphire 4870 
GraphicsGraphicsGraphicsGraphics
2x 5870, 5850, 5830, 5770 2x 3870x2, 3870 GTX Titan, GTX 480, GTX 590 GTX 285, GTX 260, 4x 9800GT, 8800GTX 
RAMHard DriveCoolingCooling
4x4GB vengeance, 2x4GB predatorX, 2x1GB OCZ DDR2 Intel X25-M 80GB LD PC-V2 SS Phase Change OCN Marksman 
CoolingCoolingOSPower
2x old tek slims (GPU) Various watercooling stuff win7, winxp AX1200 
Case
test bench / cardboard box 
  hide details  
Reply
post #125 of 203
If I understand correctly NVIDIA is doing dx11 overhead more efficiently than AMD.
post #126 of 203
Quote:
Originally Posted by Alatar View Post

I wont bother with installing the 280X again. As I told you, and showed you the variance between runs is very small no matter what the unit counts say. Not only my findings, other people have noted this as well. There's no reason to run it 10 times just to get the same results over and over again until the random avg. units numbers fit.

Even the mantle result is still very much representative. For obvious reasons listed above. The margin of error is not big enough to make the overall conclusion invalid in the slightest, even if the % might change by 1 or 2.

However since you don't want to believe that then so be it, but you can't deny that the DX11 results from both camps do show a very clear difference in DX11 CPU overhead. Hell, the Nvidia run has more average units, even though that makes next to no difference with this custom scenario.
Did you run both in something else say Heaven to ensure you are getting equal results with these clock changes to make equal? A 770 would have been a better choice.
post #127 of 203
I haven't ran that bench since 331whql but I remember the RTS setting + motion blur setting would bring my titans to its knees due to CPU overhead but that was before the developers patched the multi threading issues.
IT'S WHITE
(21 items)
 
 
CPUMotherboardGraphicsGraphics
4930k Rampage iv extreme Black Edition GTX Titan  GTX Titan 
GraphicsRAMHard DriveHard Drive
GTX Titan GskillZ Trident X Samsung EVO samsung 840 evo 
Optical DriveCoolingCoolingOS
Asus External Blue Ray drive 4x 480x60mm Swiftech MCP35x2 Win 8.1.2 
MonitorMonitorMonitorPower
AOC AOC AOC EVGA G2 1300w 
CaseMouseMouse PadAudio
STH10 RAT7 what is this?  Vali / Modi :(  
Other
Corsair HX850 
  hide details  
Reply
IT'S WHITE
(21 items)
 
 
CPUMotherboardGraphicsGraphics
4930k Rampage iv extreme Black Edition GTX Titan  GTX Titan 
GraphicsRAMHard DriveHard Drive
GTX Titan GskillZ Trident X Samsung EVO samsung 840 evo 
Optical DriveCoolingCoolingOS
Asus External Blue Ray drive 4x 480x60mm Swiftech MCP35x2 Win 8.1.2 
MonitorMonitorMonitorPower
AOC AOC AOC EVGA G2 1300w 
CaseMouseMouse PadAudio
STH10 RAT7 what is this?  Vali / Modi :(  
Other
Corsair HX850 
  hide details  
Reply
post #128 of 203
Quote:
Originally Posted by Durquavian View Post

Did you run both in something else say Heaven to ensure you are getting equal results with these clock changes to make equal? A 770 would have been a better choice.

Heaven, valley and 3dm11 would be bad ideas. They all either run too well on AMD or to well on NV. Valley and heaven to well on NV, 3dm11 p-preset too well on AMD and 3dm11 x-preset too well on Nvidia.

Firestrike extreme is really the best case scenario for as non biased as possible benchmark if you have to use just one. Also the per clock difference between the 280X and the titan matches up with what it's supposed to be.

So yeah again, sure we could nitpick and say that I should have run 10 benches from different real world games to get results accurate down to 0.1% but that's besides the overall point when the differences are so clear cut in this case (NV DX11 vs. AMD DX11).

The super accurate benches are what I leave to 3rd party reviewers who do this for a living. I was just trying to demonstrate a point. And with a huge 24% difference I think it has been demonstrated.
Quote:
Originally Posted by skupples View Post

I haven't ran that bench since 331whql but I remember the RTS setting + motion blur setting would bring my titans to its knees due to CPU overhead but that was before the developers patched the multi threading issues.

That's the whole point of the Star Swarm bench.

It's supposed to be a good mantle bench and this is achieved by applying a disgustingly draw call intensive motion blur filter. Disabling that makes the bench fly on any of our cards.
Edited by Alatar - 3/24/14 at 7:12am
 
Benching
(17 items)
 
 
CPUMotherboardGraphicsRAM
[i7 5960X @ 4.8GHz] [Rampage V Extreme] [Titan 1400MHz (1500MHz bench)] [Various] 
Hard DriveCoolingCoolingCooling
[250GB 840EVO +2x SpinpointF3 1TB RAID0] [LD PC-V2 SS Phase Change] [XSPC X2O 750 pump/res] [Monsta 360 full copper + EK XT 360 + XT 240] 
MonitorPowerCaseAudio
[Crossover 27Q LED-P 1440p+ASUS 1200p+LG 1080p] [Corsair AX1200] [Dimastech Easy v3.0] [Sennheiser HD558s] 
CPUCPUMotherboardGraphics
FX 8320, FX 8350, Phenom II x2 555BE i7 3930K, i7 860, i7 4770K, 68x Celeron D CVF, commando, 2x RIVE, Z87X-OC Asus 4870x2, Sapphire 4870 
GraphicsGraphicsGraphicsGraphics
2x 5870, 5850, 5830, 5770 2x 3870x2, 3870 GTX Titan, GTX 480, GTX 590 GTX 285, GTX 260, 4x 9800GT, 8800GTX 
RAMHard DriveCoolingCooling
4x4GB vengeance, 2x4GB predatorX, 2x1GB OCZ DDR2 Intel X25-M 80GB LD PC-V2 SS Phase Change OCN Marksman 
CoolingCoolingOSPower
2x old tek slims (GPU) Various watercooling stuff win7, winxp AX1200 
Case
test bench / cardboard box 
  hide details  
Reply
 
Benching
(17 items)
 
 
CPUMotherboardGraphicsRAM
[i7 5960X @ 4.8GHz] [Rampage V Extreme] [Titan 1400MHz (1500MHz bench)] [Various] 
Hard DriveCoolingCoolingCooling
[250GB 840EVO +2x SpinpointF3 1TB RAID0] [LD PC-V2 SS Phase Change] [XSPC X2O 750 pump/res] [Monsta 360 full copper + EK XT 360 + XT 240] 
MonitorPowerCaseAudio
[Crossover 27Q LED-P 1440p+ASUS 1200p+LG 1080p] [Corsair AX1200] [Dimastech Easy v3.0] [Sennheiser HD558s] 
CPUCPUMotherboardGraphics
FX 8320, FX 8350, Phenom II x2 555BE i7 3930K, i7 860, i7 4770K, 68x Celeron D CVF, commando, 2x RIVE, Z87X-OC Asus 4870x2, Sapphire 4870 
GraphicsGraphicsGraphicsGraphics
2x 5870, 5850, 5830, 5770 2x 3870x2, 3870 GTX Titan, GTX 480, GTX 590 GTX 285, GTX 260, 4x 9800GT, 8800GTX 
RAMHard DriveCoolingCooling
4x4GB vengeance, 2x4GB predatorX, 2x1GB OCZ DDR2 Intel X25-M 80GB LD PC-V2 SS Phase Change OCN Marksman 
CoolingCoolingOSPower
2x old tek slims (GPU) Various watercooling stuff win7, winxp AX1200 
Case
test bench / cardboard box 
  hide details  
Reply
post #129 of 203
Quote:
Originally Posted by Alatar View Post

I wont bother with installing the 280X again. As I told you, and showed you the variance between runs is very small no matter what the unit counts say. Not only my findings, other people have noted this as well. There's no reason to run it 10 times just to get the same results over and over again until the random avg. units numbers fit.

Even the mantle result is still very much representative. For obvious reasons listed above. The margin of error is not big enough to make the overall conclusion invalid in the slightest, even if the % might change by 1 or 2.

However since you don't want to believe that then so be it, but you can't deny that the DX11 results from both camps do show a very clear difference in DX11 CPU overhead. Hell, the Nvidia run has more average units, even though that makes next to no difference with this custom scenario.

I am terribly sorry but your testing has loose ends left and right and the conclusion is iffy to say the least. It's questionable your 280X vs your Titan underclocked represents stock 290X vs stock Titan (first anecdote). For all we know you could even be doing injustice to the Titan. That's a huge variable right there As for your mantle run, it had 10% more units appearing ,we aren't talking margin of error here. No matter what you think is going on with SS, it is laughable to claim that unit number isn't tied to the bench result, since it is a core element of the bench itself. It isn't measuring mosquitos on lake Pyhäjärvi in july, it's all about units and their interaction.
Mastodon Ryzen
(12 items)
 
HP Z220
(8 items)
 
 
CPUMotherboardGraphicsRAM
R7 1800X Asus Crosshair VI Hero Sapphire RX Vega 64 reference Gskill TridentZ 
Hard DriveOptical DriveCoolingOS
Pny SSD 240GB Crucial MX100 CM Nepton 280L Win 10 
MonitorPowerCaseMouse
Acer Predator XG270HU Freesync XFX 750W Pro HAF XM Logitech G502 
CPUMotherboardGraphicsCooling
i7 3770 HP Quadro K2000 HP 
OSPowerCaseMouse
Win 7  HP 400W HP CMT RAT 7 
  hide details  
Reply
Mastodon Ryzen
(12 items)
 
HP Z220
(8 items)
 
 
CPUMotherboardGraphicsRAM
R7 1800X Asus Crosshair VI Hero Sapphire RX Vega 64 reference Gskill TridentZ 
Hard DriveOptical DriveCoolingOS
Pny SSD 240GB Crucial MX100 CM Nepton 280L Win 10 
MonitorPowerCaseMouse
Acer Predator XG270HU Freesync XFX 750W Pro HAF XM Logitech G502 
CPUMotherboardGraphicsCooling
i7 3770 HP Quadro K2000 HP 
OSPowerCaseMouse
Win 7  HP 400W HP CMT RAT 7 
  hide details  
Reply
post #130 of 203
Quote:
Originally Posted by Kuivamaa View Post

I am terribly sorry but your testing has loose ends left and right and the conclusion is iffy to say the least. It's questionable your 280X vs your Titan underclocked represents stock 290X vs stock Titan (first anecdote). For all we know you could even be doing injustice to the Titan. That's a huge variable right there As for your mantle run, it had 10% more units appearing ,we aren't talking margin of error here. No matter what you think is going on with SS, it is laughable to claim that unit number isn't tied to the bench result, since it is a core element of the bench itself. It isn't measuring mosquitos on lake Pyhäjärvi in july, it's all about units and their interaction.

If you've got better data, I'd love to see it. As it stands now, I'll take what I can get, with acknowledgements for the imperfections. How does shooting down people who are trying to provide data and replacing it with nothing help anyone?
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [pcper] NVIDIA Talks DX12, DX11 Efficiency Improvements