Overclock.net › Forums › Industry News › Software News › [Guru3d] Guru3D Releases "Catzilla" benchmark
New Posts  All Forums:Forum Nav:

[Guru3d] Guru3D Releases "Catzilla" benchmark - Page 28

post #271 of 333
Quote:
Originally Posted by nismo_usaf View Post

Maybe its CPU's throwing the scores off.

nope

plenty of guys like me and the guy above you with 4.8 ghz + SB-E chips that are getting worse overall scores then people with moderately clocked i5's even though they have much worse scores in the physics portion of the test. The Nvidia systems are only winning soundly in the first part of the test; all my other individual scores are usually better then 680 systems but the "Hardware" portion which makes up the bulk of the score always seems to be about 1k points ahead
Poison
(13 items)
 
  
CPUMotherboardGraphicsRAM
3820 @ 4.7 Sabertooth x79 Powercolor 6870x2 G Skill 2133 8gb 
Hard DriveCoolingOSMonitor
Crucial m4 Noctua DH14 win7 64 ultimate HP 2511x 
KeyboardPowerMouseMouse Pad
microsoft Sidewinder Enermax Modu 87+ 900 Asus GX800 none 
Audio
Yamaha RX-V955, B&W bookshelves and Infinity PS-8 
  hide details  
Reply
Poison
(13 items)
 
  
CPUMotherboardGraphicsRAM
3820 @ 4.7 Sabertooth x79 Powercolor 6870x2 G Skill 2133 8gb 
Hard DriveCoolingOSMonitor
Crucial m4 Noctua DH14 win7 64 ultimate HP 2511x 
KeyboardPowerMouseMouse Pad
microsoft Sidewinder Enermax Modu 87+ 900 Asus GX800 none 
Audio
Yamaha RX-V955, B&W bookshelves and Infinity PS-8 
  hide details  
Reply
post #272 of 333
3DMark Vantage had hardware PhysX enabled by default originally, and that caused some people to get their panties in a knot as well. It's possible that the same is happening here, but I guess we'll have to wait and see.
    
CPUMotherboardGraphicsRAM
i7 920 D0 MSI X58 Pro-E GTX 560 Ti 448 3x2GB G.Skill DDR3-1333 9-9-9-24 
Hard DriveHard DriveOptical DriveOS
840 Pro Caviar Black LG BD-ROM Windows 8.1 Pro x64 
MonitorMonitorKeyboardPower
Dell U2713HM Dell U2311H Turbo-Trak (Google it :D) Corsair HX-520 
CaseMouseMouse PadAudio
CM690 Mionix Avior 7000 Everglide Titan AKG K 242 HD 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
i7 920 D0 MSI X58 Pro-E GTX 560 Ti 448 3x2GB G.Skill DDR3-1333 9-9-9-24 
Hard DriveHard DriveOptical DriveOS
840 Pro Caviar Black LG BD-ROM Windows 8.1 Pro x64 
MonitorMonitorKeyboardPower
Dell U2713HM Dell U2311H Turbo-Trak (Google it :D) Corsair HX-520 
CaseMouseMouse PadAudio
CM690 Mionix Avior 7000 Everglide Titan AKG K 242 HD 
  hide details  
Reply
post #273 of 333
Quote:
Originally Posted by Acefire View Post

Nvidia biased to the max. Not a legit benchmark by any stretch.

Quote:
Originally Posted by Acefire View Post

Quote:
Originally Posted by nismo_usaf View Post

Excuse me Sir, but where is your benchmark? Seems as 95% of your post are "red team biased".

It doesn't seem fishy to you that this is the only benchmark where Nvidia wins across the board? Stock 670 gets amazing score whilst 7970 @ 1.3ghz pulls in like 8k.

Junk benchmark.

Quote:
Originally Posted by Acefire View Post

We'll see. Hopefully the scores make some more sense by then. IT you look at the leaderboard there are people with 1 680 that have a score of like 28,000

I hope you're not as stupid as you sound.

Each GPU type has a different architecture. Back in the VLIW days, AMD were counting on 4+1 vector calculations to gain top spot, then they changed to GCN which is still using as much IPC as can be. NVIDIA's take is all about thread scheduling, and simple warp placement within the SMs.

Each physics engine has a particular way of calculating things - some benefit VLIW, others GCN, and some benefit Fermi and Kepler. Different engines will send different calculations to the GPU to be processed.

If you take any one engine in particular and play the same actions on a system, then replace the GPU with a different architecture, they will all have the same commands issued to them through the DirectX/OpenGL/DirectCompute protocols, but it is up to the silicon to decide how to process them. It boils down to scheduling, to IPC, size of caches, texture buffers, among many other things. If the calculation order or process benefits one architecture, then that is the fault of the PHYSICS ENGINE, not the developers. Most developers will use the physics engine they feel most comfortable with or that accurately reflects what they want to do, not out a sense of loyalty for brand XYZ or in the interests of fairness.

Get off your high horse and actually understand the role GPU architecture plays in this.
post #274 of 333
Quote:
Originally Posted by borandi View Post

Quote:
Originally Posted by Acefire View Post

Nvidia biased to the max. Not a legit benchmark by any stretch.

Quote:
Originally Posted by Acefire View Post

Quote:
Originally Posted by nismo_usaf View Post

Excuse me Sir, but where is your benchmark? Seems as 95% of your post are "red team biased".

It doesn't seem fishy to you that this is the only benchmark where Nvidia wins across the board? Stock 670 gets amazing score whilst 7970 @ 1.3ghz pulls in like 8k.

Junk benchmark.

Quote:
Originally Posted by Acefire View Post

We'll see. Hopefully the scores make some more sense by then. IT you look at the leaderboard there are people with 1 680 that have a score of like 28,000

I hope you're not as stupid as you sound.

Each GPU type has a different architecture. Back in the VLIW days, AMD were counting on 4+1 vector calculations to gain top spot, then they changed to GCN which is still using as much IPC as can be. NVIDIA's take is all about thread scheduling, and simple warp placement within the SMs.

Each physics engine has a particular way of calculating things - some benefit VLIW, others GCN, and some benefit Fermi and Kepler. Different engines will send different calculations to the GPU to be processed.

If you take any one engine in particular and play the same actions on a system, then replace the GPU with a different architecture, they will all have the same commands issued to them through the DirectX/OpenGL/DirectCompute protocols, but it is up to the silicon to decide how to process them. It boils down to scheduling, to IPC, size of caches, texture buffers, among many other things. If the calculation order or process benefits one architecture, then that is the fault of the PHYSICS ENGINE, not the developers. Most developers will use the physics engine they feel most comfortable with or that accurately reflects what they want to do, not out a sense of loyalty for brand XYZ or in the interests of fairness.

Get off your high horse and actually understand the role GPU architecture plays in this.

The benchmark does not show the relative difference between the GPUs that is seen both in the real world and in other benchmarks.

It also thinks a stock 2500k should get the same physics score as a 8320 at 5Ghz, and it does use (or tries to use) all 8 threads, by testing done between britishbob and myself.

The benchmark is biased. Get off your high horse and accept it.
Forge
(17 items)
 
Forge-LT
(7 items)
 
 
CPUMotherboardGraphicsRAM
Intel i7-5960X (4.625Ghz) ASUS X99-DELUXE/U3.1 2x AMD R9-FuryX 64GB Corsair Dominator Platinum (3000Mhz 8x8GB) 
Hard DriveCoolingOSMonitor
Samsung 950 Pro NVMe 512GB EK Predator 240 Windows 10 Enterprise x64 2x Acer XR341CK 
KeyboardPowerCaseMouse
Corsair Vengeance K70 RGB Corsair AX1200 Thermaltake X9 Corsair Vengeance M65 RGB 
AudioAudioAudioAudio
Sennheiser HD700 Objective2+ODAC REV B Sound Blaster ZxR Audio Technica AT2020 
Audio
30ART Mic Tube Amp 
CPUMotherboardGraphicsRAM
i7-4720HQ UX501JW-UB71T GTX 960m 16GB 1600 9-9-9-27 
Hard DriveOSMonitor
512GB PCI-e SSD Windows 10 Pro 4k IPS 
  hide details  
Reply
Forge
(17 items)
 
Forge-LT
(7 items)
 
 
CPUMotherboardGraphicsRAM
Intel i7-5960X (4.625Ghz) ASUS X99-DELUXE/U3.1 2x AMD R9-FuryX 64GB Corsair Dominator Platinum (3000Mhz 8x8GB) 
Hard DriveCoolingOSMonitor
Samsung 950 Pro NVMe 512GB EK Predator 240 Windows 10 Enterprise x64 2x Acer XR341CK 
KeyboardPowerCaseMouse
Corsair Vengeance K70 RGB Corsair AX1200 Thermaltake X9 Corsair Vengeance M65 RGB 
AudioAudioAudioAudio
Sennheiser HD700 Objective2+ODAC REV B Sound Blaster ZxR Audio Technica AT2020 
Audio
30ART Mic Tube Amp 
CPUMotherboardGraphicsRAM
i7-4720HQ UX501JW-UB71T GTX 960m 16GB 1600 9-9-9-27 
Hard DriveOSMonitor
512GB PCI-e SSD Windows 10 Pro 4k IPS 
  hide details  
Reply
post #275 of 333
Quote:
Originally Posted by KyadCK View Post

The benchmark does not show the relative difference between the GPUs that is seen both in the real world and in other benchmarks.

Fire up some code that compiles to a naturally easy shortcut in one architecture and the other architecture will seem to struggle. You don't seem to get this. If I write OpenCL code in a normal way and it maps nicer to AMD than NVIDIA, does that show the differences between the GPU? No.

There is no single-algorithm panacea that maps the same to each vendor's hardware. If there is a score difference, then that is representative of the how each architecture deals with the mathematics of the engine.

When you build a bit of silicon, you determine what are the easy wins from the last silicon (more cache, more schedulers, better rasterisers) and hope it works for the next gen. It's not the fault of the programmer if the silicon makers decided to go after one area they thought would benefit than something else which just happens to be what you are testing.
Quote:
It also thinks a stock 2500k should get the same physics score as a 8320 at 5Ghz, and it does use (or tries to use) all 8 threads, by testing done between britishbob and myself.

You do understand how a 8320 works, right? Four modules, and each module has two INT units and one FP unit for calculation. That means that in floating point arithmetic, it acts like an 4-core, and in integer arithmetic, it acts like an 8-core. AMD went this way on the understanding that OS logic is often driven by int loops. But most physics and mathematics is driven by floating point mathematics, like physics in games. Only the highly optimised algorithms solely use integer arithmetic.

It has been shown repeatedly that the 2500K in floating point arithmetic is 30% faster in single threaded mode than an 8320. This is shown in the AnandTech 3DPM test which is memory independent and uses solely FP mathematics (using an 8150):



Even when using ALL 8 THREADS of that 8120, the i5-2500K still beat it:



Primary physics, such as those used in games, are only slightly multithreaded. They involve memory fences and synchronization gaps which slow down the system, and if the scheduler of the processor is not up to the task, then context thread switching performance literally sucks. This is why only secondary physics, which is allowed to make mistakes because it doesn't influence the game one iota, can be sent to GPUs.

Unless you can show me that the physics operations in the engine that the benchmark uses contains primarily integer mathematics (highly unlikely if you have EVER worked with a physics engine), I would happily believe the results are realistic and real-world, regardless if you think they are biased.
post #276 of 333
that's a pretty awesome animation thumb.gif
Current Rig
(10 items)
 
Old
(11 items)
 
2014
(14 items)
 
CPUMotherboardGraphicsRAM
AMD Athlon 64 X2 3800+ Foxxconn 6150BK8MC 256MB ATI RADEON X800 PCIE 2.5 GB DDR400 
Hard DriveCoolingOSMonitor
80 GB Western Digital Thermaltake GunMet Orb MS Windows 7 Ultimate 64-bit Acer S230HL 
PowerCase
Allied SL-8360BTX ATX 350w Antec 300 
CPUMotherboardGraphicsRAM
AMD Athlon 64 3000+ K8 Combo-Z 256MB ATI Radeon 9550 / X1050 Series 1 GB 
Hard DriveOptical DriveOSMonitor
80 GB Western Digital OPTORITE CD-RW MS Windows XP Professional 32-bit SP3 Acer S230HL 23-Inch 
Power
Orion YH-480w 
CPUMotherboardGraphicsRAM
Intel i5-4670K MSI Z87-G45 MSI GTX 760 TF 2GD5 8GB G.SKILL Trident X DDR3-2400 
Hard DriveCoolingOSMonitor
Seagate Barracuda 500GB Cooler Master Hyper 212 EVO Microsoft Windows 10 Acer S230HL 23" 
KeyboardPowerCaseMouse
Logitech G105 SeaSonic M12II 620w 80+ Bronze Antec 300 Razer Deathadder 2013 
  hide details  
Reply
Current Rig
(10 items)
 
Old
(11 items)
 
2014
(14 items)
 
CPUMotherboardGraphicsRAM
AMD Athlon 64 X2 3800+ Foxxconn 6150BK8MC 256MB ATI RADEON X800 PCIE 2.5 GB DDR400 
Hard DriveCoolingOSMonitor
80 GB Western Digital Thermaltake GunMet Orb MS Windows 7 Ultimate 64-bit Acer S230HL 
PowerCase
Allied SL-8360BTX ATX 350w Antec 300 
CPUMotherboardGraphicsRAM
AMD Athlon 64 3000+ K8 Combo-Z 256MB ATI Radeon 9550 / X1050 Series 1 GB 
Hard DriveOptical DriveOSMonitor
80 GB Western Digital OPTORITE CD-RW MS Windows XP Professional 32-bit SP3 Acer S230HL 23-Inch 
Power
Orion YH-480w 
CPUMotherboardGraphicsRAM
Intel i5-4670K MSI Z87-G45 MSI GTX 760 TF 2GD5 8GB G.SKILL Trident X DDR3-2400 
Hard DriveCoolingOSMonitor
Seagate Barracuda 500GB Cooler Master Hyper 212 EVO Microsoft Windows 10 Acer S230HL 23" 
KeyboardPowerCaseMouse
Logitech G105 SeaSonic M12II 620w 80+ Bronze Antec 300 Razer Deathadder 2013 
  hide details  
Reply
post #277 of 333
Quote:
Originally Posted by borandi View Post

Quote:
It also thinks a stock 2500k should get the same physics score as a 8320 at 5Ghz, and it does use (or tries to use) all 8 threads, by testing done between britishbob and myself.

You do understand how a 8320 works, right? Four modules, and each module has two INT units and one FP unit for calculation. That means that in floating point arithmetic, it acts like an 4-core, and in integer arithmetic, it acts like an 8-core. AMD went this way on the understanding that OS logic is often driven by int loops. But most physics and mathematics is driven by floating point mathematics, like physics in games. Only the highly optimised algorithms solely use integer arithmetic.

It has been shown repeatedly that the 2500K in floating point arithmetic is 30% faster in single threaded mode than an 8320. This is shown in the AnandTech 3DPM test which is memory independent and uses solely FP mathematics (using an 8150):

8320. At 5Ghz. which is 43% faster then stock. Where's your math now?

Also, at least try to find a modern test. BD is not PD.

Actually, you know what? Lets use your multithreaded graph. The 8150 is just behind the 2500k, yes? Good. Now add my 7% IPC advantage for having PD instead. And now add my 39% clock speed bonus on top of that. I should be coming out with a score of 505 assuming perfect scaling, but even without that I still absolutely destroy the 2500k at stock, which is what my CPU ties with according to Catzilla.

Still think it's "realistic"?
Edited by KyadCK - 1/1/13 at 9:10am
Forge
(17 items)
 
Forge-LT
(7 items)
 
 
CPUMotherboardGraphicsRAM
Intel i7-5960X (4.625Ghz) ASUS X99-DELUXE/U3.1 2x AMD R9-FuryX 64GB Corsair Dominator Platinum (3000Mhz 8x8GB) 
Hard DriveCoolingOSMonitor
Samsung 950 Pro NVMe 512GB EK Predator 240 Windows 10 Enterprise x64 2x Acer XR341CK 
KeyboardPowerCaseMouse
Corsair Vengeance K70 RGB Corsair AX1200 Thermaltake X9 Corsair Vengeance M65 RGB 
AudioAudioAudioAudio
Sennheiser HD700 Objective2+ODAC REV B Sound Blaster ZxR Audio Technica AT2020 
Audio
30ART Mic Tube Amp 
CPUMotherboardGraphicsRAM
i7-4720HQ UX501JW-UB71T GTX 960m 16GB 1600 9-9-9-27 
Hard DriveOSMonitor
512GB PCI-e SSD Windows 10 Pro 4k IPS 
  hide details  
Reply
Forge
(17 items)
 
Forge-LT
(7 items)
 
 
CPUMotherboardGraphicsRAM
Intel i7-5960X (4.625Ghz) ASUS X99-DELUXE/U3.1 2x AMD R9-FuryX 64GB Corsair Dominator Platinum (3000Mhz 8x8GB) 
Hard DriveCoolingOSMonitor
Samsung 950 Pro NVMe 512GB EK Predator 240 Windows 10 Enterprise x64 2x Acer XR341CK 
KeyboardPowerCaseMouse
Corsair Vengeance K70 RGB Corsair AX1200 Thermaltake X9 Corsair Vengeance M65 RGB 
AudioAudioAudioAudio
Sennheiser HD700 Objective2+ODAC REV B Sound Blaster ZxR Audio Technica AT2020 
Audio
30ART Mic Tube Amp 
CPUMotherboardGraphicsRAM
i7-4720HQ UX501JW-UB71T GTX 960m 16GB 1600 9-9-9-27 
Hard DriveOSMonitor
512GB PCI-e SSD Windows 10 Pro 4k IPS 
  hide details  
Reply
post #278 of 333
I score 10300 on Tiger Preset on sig rig. Very disappointed in 7970 CF on this bench.
X79-GCN
(22 items)
 
  
CPUMotherboardGraphicsRAM
Intel 3930K 4.5GHz HT GIGABYTE GA-X79-UP4 AMD R9-290X GEil Evo Potenza DDR3 2400MHz CL10 (4x4GB) 
Hard DriveCoolingCoolingCooling
Samsung 840 Pro 120GB EK Supremacy (CPU) NF F12's P/P (360 Rad)  NF A14's (420 Rad)  
CoolingCoolingCoolingCooling
XSPC Chrome Compression Fittings EK RES X3 150 Primochill PremoFlex Advanced LRT Clear 1/2 ID EK-FC (R9 290X) 
CoolingCoolingCoolingOS
EK D5 Vario Top-X  Phobya G-Changer V2 360mm Phobya G-Changer V2 420mm Win 10 x64 Pro 
MonitorKeyboardPowerCase
BenQ XR3501 35" Curved Corsair Vengeance K90 Seasonic X-1250 Gold (v2) Corsair 900D 
MouseAudio
Logitech G400s Senn HD 598 
  hide details  
Reply
X79-GCN
(22 items)
 
  
CPUMotherboardGraphicsRAM
Intel 3930K 4.5GHz HT GIGABYTE GA-X79-UP4 AMD R9-290X GEil Evo Potenza DDR3 2400MHz CL10 (4x4GB) 
Hard DriveCoolingCoolingCooling
Samsung 840 Pro 120GB EK Supremacy (CPU) NF F12's P/P (360 Rad)  NF A14's (420 Rad)  
CoolingCoolingCoolingCooling
XSPC Chrome Compression Fittings EK RES X3 150 Primochill PremoFlex Advanced LRT Clear 1/2 ID EK-FC (R9 290X) 
CoolingCoolingCoolingOS
EK D5 Vario Top-X  Phobya G-Changer V2 360mm Phobya G-Changer V2 420mm Win 10 x64 Pro 
MonitorKeyboardPowerCase
BenQ XR3501 35" Curved Corsair Vengeance K90 Seasonic X-1250 Gold (v2) Corsair 900D 
MouseAudio
Logitech G400s Senn HD 598 
  hide details  
Reply
post #279 of 333
Quote:
Originally Posted by skyn3t View Post

After some bios issues i manager to fix my mod and have my fixed clock up and running so my card stays on top clock all the time smile.gif .
PS: For those have SLI , before flash the cards make sure you remove the SLI BRIDGE before any flash.

I've flashed mine around 10 times and I've never removed the SLI bridge.
Mostly For Gaming
(15 items)
 
  
CPUMotherboardGraphicsRAM
i7-4820K EK cooled ASUS x79 680, EK cooled G.skill 16GB 
Hard DriveOptical DriveCoolingOS
512GB 840 Pro  ASUS DVD Burner 3 Loops which consist of 7 Waterblocks, 4 Radia... 8.1pro 64bit 
MonitorKeyboardPowerCase
ZR30w S-IPS 2560x1600 Ɠ19 SILVERSTONE ST 1500W Custom 
MouseMouse PadAudio
Ɠ500 Ɍocketfish SENNHEISER 
  hide details  
Reply
Mostly For Gaming
(15 items)
 
  
CPUMotherboardGraphicsRAM
i7-4820K EK cooled ASUS x79 680, EK cooled G.skill 16GB 
Hard DriveOptical DriveCoolingOS
512GB 840 Pro  ASUS DVD Burner 3 Loops which consist of 7 Waterblocks, 4 Radia... 8.1pro 64bit 
MonitorKeyboardPowerCase
ZR30w S-IPS 2560x1600 Ɠ19 SILVERSTONE ST 1500W Custom 
MouseMouse PadAudio
Ɠ500 Ɍocketfish SENNHEISER 
  hide details  
Reply
post #280 of 333
Does my score look normal for stock 7970's on stock speeds?
Tiger Preset

Cheers.
X79-GCN
(22 items)
 
  
CPUMotherboardGraphicsRAM
Intel 3930K 4.5GHz HT GIGABYTE GA-X79-UP4 AMD R9-290X GEil Evo Potenza DDR3 2400MHz CL10 (4x4GB) 
Hard DriveCoolingCoolingCooling
Samsung 840 Pro 120GB EK Supremacy (CPU) NF F12's P/P (360 Rad)  NF A14's (420 Rad)  
CoolingCoolingCoolingCooling
XSPC Chrome Compression Fittings EK RES X3 150 Primochill PremoFlex Advanced LRT Clear 1/2 ID EK-FC (R9 290X) 
CoolingCoolingCoolingOS
EK D5 Vario Top-X  Phobya G-Changer V2 360mm Phobya G-Changer V2 420mm Win 10 x64 Pro 
MonitorKeyboardPowerCase
BenQ XR3501 35" Curved Corsair Vengeance K90 Seasonic X-1250 Gold (v2) Corsair 900D 
MouseAudio
Logitech G400s Senn HD 598 
  hide details  
Reply
X79-GCN
(22 items)
 
  
CPUMotherboardGraphicsRAM
Intel 3930K 4.5GHz HT GIGABYTE GA-X79-UP4 AMD R9-290X GEil Evo Potenza DDR3 2400MHz CL10 (4x4GB) 
Hard DriveCoolingCoolingCooling
Samsung 840 Pro 120GB EK Supremacy (CPU) NF F12's P/P (360 Rad)  NF A14's (420 Rad)  
CoolingCoolingCoolingCooling
XSPC Chrome Compression Fittings EK RES X3 150 Primochill PremoFlex Advanced LRT Clear 1/2 ID EK-FC (R9 290X) 
CoolingCoolingCoolingOS
EK D5 Vario Top-X  Phobya G-Changer V2 360mm Phobya G-Changer V2 420mm Win 10 x64 Pro 
MonitorKeyboardPowerCase
BenQ XR3501 35" Curved Corsair Vengeance K90 Seasonic X-1250 Gold (v2) Corsair 900D 
MouseAudio
Logitech G400s Senn HD 598 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Software News
Overclock.net › Forums › Industry News › Software News › [Guru3d] Guru3D Releases "Catzilla" benchmark