Overclock.net › Forums › AMD › AMD CPUs › Is the AMD FX series being correctly benchmarked?
New Posts  All Forums:Forum Nav:

Is the AMD FX series being correctly benchmarked?

post #1 of 13
Thread Starter 
I was talking about the benchmarks of the AMD FX series with a friend and he said that's it's possible they didn't benchmark it correctly.

He claims that the chip is seen as 2 dual quads on the system.

Anyone else want to help me understand if he's right and if he is, would that affect benchmark results?
Phoenix
(16 items)
 
HP Envy 14 2011
(13 items)
 
 
CPUMotherboardGraphicsRAM
Intel Core i7 740QM Hewlett-Packard 1437 AMD Radeon HD 6630 GPU with 1GB of VRAM 6.00 GB Dual-Channel DDR3 
Hard DriveOSMonitor
750GB Western Digital MS Windows 7 Home Premium 64-bit SP1 14.5in LCD 
  hide details  
Reply
Phoenix
(16 items)
 
HP Envy 14 2011
(13 items)
 
 
CPUMotherboardGraphicsRAM
Intel Core i7 740QM Hewlett-Packard 1437 AMD Radeon HD 6630 GPU with 1GB of VRAM 6.00 GB Dual-Channel DDR3 
Hard DriveOSMonitor
750GB Western Digital MS Windows 7 Home Premium 64-bit SP1 14.5in LCD 
  hide details  
Reply
post #2 of 13
they were bench marked correctly.

They just turned off all the cores (units) bar 2 so it was effectively running as a dual core like most people do when they benchmark either AMD/intel
post #3 of 13
Quote:
Originally Posted by Drybones5 View Post
I was talking about the benchmarks of the AMD FX series with a friend and he said that's it's possible they didn't benchmark it correctly.

He claims that the chip is seen as 2 dual quads on the system.

Anyone else want to help me understand if he's right and if he is, would that affect benchmark results?
He was just making excuses for it. It's happens a lot
post #4 of 13
Quote:
Originally Posted by Don Karnage View Post
He was just making excuses for it. It's happens a lot
Depends on the compiler the bench is based on too.

Here are some good reads on this:
http://www.ftc.gov/opa/2010/08/intel.shtm (the results)
http://www.ftc.gov/opa/2009/12/intel.shtm
http://blogs.amd.com/work/2010/01/22...md-processors/

"....disclose to software developers that Intel computer compilers discriminate between Intel chips and non-Intel chips, and that they may not register all the features of non-Intel chips. Intel also will have to reimburse all software vendors who want to recompile their software using a non-Intel compiler."
This is from the FTC settlement, so if the software developer knows about this but chooses not to modify the compiler, then it still affects non-Intel CPUs, but Intel is not liable for that action. Food for thought.


Also I wish that someone with access to the SPEC CPU2006 benchmark would run it on BD.
Edited by Obakemono - 10/27/11 at 5:07pm
BULLDOZER!!!
(14 items)
 
  
CPUMotherboardGraphicsRAM
8120FX Gigabyte GA-990FXA-UD7 2ea MSI r9-270x 16gb G-Skill Snipers 
Hard DriveOptical DriveCoolingOS
WD 3x1tb greens, 1x1tb black Adata S599 120gb LG DVDR/RW, LG Blueray CM V8 Win7 64bit 
MonitorKeyboardPowerCase
LG2762D Microsoft Antec TPN-750 CM HAF 932 
MouseMouse Pad
Microsoft X5 Generic 
  hide details  
Reply
BULLDOZER!!!
(14 items)
 
  
CPUMotherboardGraphicsRAM
8120FX Gigabyte GA-990FXA-UD7 2ea MSI r9-270x 16gb G-Skill Snipers 
Hard DriveOptical DriveCoolingOS
WD 3x1tb greens, 1x1tb black Adata S599 120gb LG DVDR/RW, LG Blueray CM V8 Win7 64bit 
MonitorKeyboardPowerCase
LG2762D Microsoft Antec TPN-750 CM HAF 932 
MouseMouse Pad
Microsoft X5 Generic 
  hide details  
Reply
post #5 of 13
Quote:
Originally Posted by Obakemono View Post
Depends on the compiler the bench is based on too.

Here are some good reads on this:
http://www.ftc.gov/opa/2010/08/intel.shtm (the results)
http://www.ftc.gov/opa/2009/12/intel.shtm
http://blogs.amd.com/work/2010/01/22...md-processors/

"....disclose to software developers that Intel computer compilers discriminate between Intel chips and non-Intel chips, and that they may not register all the features of non-Intel chips. Intel also will have to reimburse all software vendors who want to recompile their software using a non-Intel compiler."
This is from the FTC settlement, so if the software developer knows about this but chooses not to modify the compiler, then it still affects non-Intel CPUs, but Intel is not liable for that action. Food for thought.
Doesn't explain its identical poor performance in Linux benchmarks. I highly doubt there is a conspiracy against AMD when it comes to benchmarks, and even if there was I doubt Intel is creating biases against it in the Linux world.

Also citing AMD's marketing people isn't really informative.
LV-426
(15 items)
 
  
CPUMotherboardGraphicsRAM
Intel 3930k Asrock x79 Extreme6 Gigabyte 6950 unlocked 920/1375 16gb G.Skill DDR3 2133 mhz 
Hard DriveOptical DriveCoolingOS
OCZ Agility 2 60 gb + 7.5 TBs of storage Pioneer DVD-RW Cooler Master 212+ Evo Windows 7 x64 Ultimate 
MonitorKeyboardPowerCase
ViewSonic 22" 1080p IBM Model M Corsair HX 1050w Cooler Master HAF X 
MouseAudio
Razer Lachesis 4000 dpi Creative XtremeMusic 
  hide details  
Reply
LV-426
(15 items)
 
  
CPUMotherboardGraphicsRAM
Intel 3930k Asrock x79 Extreme6 Gigabyte 6950 unlocked 920/1375 16gb G.Skill DDR3 2133 mhz 
Hard DriveOptical DriveCoolingOS
OCZ Agility 2 60 gb + 7.5 TBs of storage Pioneer DVD-RW Cooler Master 212+ Evo Windows 7 x64 Ultimate 
MonitorKeyboardPowerCase
ViewSonic 22" 1080p IBM Model M Corsair HX 1050w Cooler Master HAF X 
MouseAudio
Razer Lachesis 4000 dpi Creative XtremeMusic 
  hide details  
Reply
post #6 of 13
Not quite as bleak on Linux.

http://www.phoronix.com/scan.php?pag...ulldozer&num=1

Main thing is FX-8150 should be around 2500K price and would become a decent buy for many MT workloads. Very similar to Phenom II x6 competing against first gen i5 series. Will be interesting to see how Interlagos BD launch turns out, seems that will be the only way to move this first gen chip in volume and make some profit.
Everyday PC
(13 items)
 
  
CPUMotherboardGraphicsRAM
i5-3470 (+4 Turbo bins) Biostar T77XE3 Radeon 7950 16GB DDR3 1600 
Hard DriveOSMonitorKeyboard
Samsung F3 1TB Windows 7 Pro x64 Samsung 24" TN MS Natural Elite 4000 
PowerCaseMouse
Antec HPG 900W Lian-li K57W Logitech G500 
  hide details  
Reply
Everyday PC
(13 items)
 
  
CPUMotherboardGraphicsRAM
i5-3470 (+4 Turbo bins) Biostar T77XE3 Radeon 7950 16GB DDR3 1600 
Hard DriveOSMonitorKeyboard
Samsung F3 1TB Windows 7 Pro x64 Samsung 24" TN MS Natural Elite 4000 
PowerCaseMouse
Antec HPG 900W Lian-li K57W Logitech G500 
  hide details  
Reply
post #7 of 13
Quote:
Originally Posted by PvtHudson View Post
Doesn't explain its identical poor performance in Linux benchmarks. I highly doubt there is a conspiracy against AMD when it comes to benchmarks, and even if there was I doubt Intel is creating biases against it in the Linux world.

Also citing AMD's marketing people isn't really informative.
Yes AMD got 1billion + dollars out of Intel because they were playing fair

The compiler bias is still there it has not been fixed.

Some of the opensource librarys have been built using the Intel Compiler. Like AMD's very own Kernel Math Library.

Its a very real problem.

The fustrating part is figuring out what the bias is and how big or small it is in reality. Phoronix got test results from fx8150 and its was trading places with the i2500/i2600 fiarly regularly in most application. Ahead in some behind in others but far more balanced. There was certainly a decline in single threaded performance vrs Ph2.

However the scope of the disparity and the distance in the Phoronix benchmarks is vastly smaller in the Linux benchmarks.

I think AMD just missed the target a bit, found some hard to rectify bottle neck late in the design and had to push forward. For the new Piledriver core to be so close, they had to know that they weren't going to be competitive enough. Combined with GloFo not delivering on the process node to get the clockspeeds and the yeilds they needed with good power consumption, it all add's up to a perfect storm.

Lets see how this design matures going forward.
post #8 of 13
Quote:
Originally Posted by Don Karnage View Post
He was just making excuses for it. It's happens a lot
No one made excuses, it was only someone thoughts.
    
CPUMotherboardGraphicsRAM
8350/4.8 Asus Crosshair v formula z Sapphire Vapor x 7950 Crucial BalisticTactical tracer-1866/9-9-9-24 O... 
Hard DriveHard DriveOptical DriveCooling
Crucial m4 128gb Western Digital Caviar Black 2 TB SATA III 7200... Asus Burner Corsair H100  
OSMonitorKeyboardPower
windows 7 ultimate 64 bit Asus ve258q Logitech G19 seasonic series x-1250  
CaseMouseMouse PadAudio
InWin Dragon Rider Logitech G 500 thermaltake lighted logitech 5.1 506 
Other
SB 3D recon sound card 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
8350/4.8 Asus Crosshair v formula z Sapphire Vapor x 7950 Crucial BalisticTactical tracer-1866/9-9-9-24 O... 
Hard DriveHard DriveOptical DriveCooling
Crucial m4 128gb Western Digital Caviar Black 2 TB SATA III 7200... Asus Burner Corsair H100  
OSMonitorKeyboardPower
windows 7 ultimate 64 bit Asus ve258q Logitech G19 seasonic series x-1250  
CaseMouseMouse PadAudio
InWin Dragon Rider Logitech G 500 thermaltake lighted logitech 5.1 506 
Other
SB 3D recon sound card 
  hide details  
Reply
post #9 of 13
Thread Starter 
Quote:
Originally Posted by Vesku View Post
Not quite as bleak on Linux.

http://www.phoronix.com/scan.php?pag...ulldozer&num=1

Main thing is FX-8150 should be around 2500K price and would become a decent buy for many MT workloads. Very similar to Phenom II x6 competing against first gen i5 series. Will be interesting to see how Interlagos BD launch turns out, seems that will be the only way to move this first gen chip in volume and make some profit.
http://www.phoronix.com/scan.php?pag...lldozer&num=13

I read through all that. It seems like the Bulldozer series isn't not even as nearly bad as we thought. It does video encoding very well.

Plus this was on un-optimized and incomplete kernels for the cpu.

In addition, they said we won't see the true ability of the FX series till Windows 8. I'm still questioning if I want to wait till the next release of the FX series to wait for more hardware optimization.
Phoenix
(16 items)
 
HP Envy 14 2011
(13 items)
 
 
CPUMotherboardGraphicsRAM
Intel Core i7 740QM Hewlett-Packard 1437 AMD Radeon HD 6630 GPU with 1GB of VRAM 6.00 GB Dual-Channel DDR3 
Hard DriveOSMonitor
750GB Western Digital MS Windows 7 Home Premium 64-bit SP1 14.5in LCD 
  hide details  
Reply
Phoenix
(16 items)
 
HP Envy 14 2011
(13 items)
 
 
CPUMotherboardGraphicsRAM
Intel Core i7 740QM Hewlett-Packard 1437 AMD Radeon HD 6630 GPU with 1GB of VRAM 6.00 GB Dual-Channel DDR3 
Hard DriveOSMonitor
750GB Western Digital MS Windows 7 Home Premium 64-bit SP1 14.5in LCD 
  hide details  
Reply
post #10 of 13
That's another major thorn in the FX launch, by the time the software is ready for it they will probably have released at least a revision if not Piledriver. No reason not to just pass on current FX and check on it in 6-9 months.
Everyday PC
(13 items)
 
  
CPUMotherboardGraphicsRAM
i5-3470 (+4 Turbo bins) Biostar T77XE3 Radeon 7950 16GB DDR3 1600 
Hard DriveOSMonitorKeyboard
Samsung F3 1TB Windows 7 Pro x64 Samsung 24" TN MS Natural Elite 4000 
PowerCaseMouse
Antec HPG 900W Lian-li K57W Logitech G500 
  hide details  
Reply
Everyday PC
(13 items)
 
  
CPUMotherboardGraphicsRAM
i5-3470 (+4 Turbo bins) Biostar T77XE3 Radeon 7950 16GB DDR3 1600 
Hard DriveOSMonitorKeyboard
Samsung F3 1TB Windows 7 Pro x64 Samsung 24" TN MS Natural Elite 4000 
PowerCaseMouse
Antec HPG 900W Lian-li K57W Logitech G500 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: AMD CPUs
Overclock.net › Forums › AMD › AMD CPUs › Is the AMD FX series being correctly benchmarked?