Overclock.net › Forums › AMD › AMD CPUs › Questions about the FX 8350?
New Posts  All Forums:Forum Nav:

Questions about the FX 8350? - Page 4

post #31 of 402
Quote:
Originally Posted by mdocod View Post

I'm sorry I think that's giving the PD too much credit. The FX-8350 can indeed be much faster than an i5-2500K in non-real-time workloads, but in terms of gaming (real-time) that this thread is concerned about, Sandy Bridge has a more robust core architecture and tends to clock into the near 5ghz class like PD, which gives it the edge in these less-threaded, difficult to achieve saturation workloads. PD/BD was originally slated to be in competition with Nahalem, and it has very similar performance characteristics. An FX-83XX@5ghz is almost a dead even match-up to an i7-970(hex core)@4ghz in all sorts of workloads. (real-time or otherwise).

Sandy Bridge changed everything with a decent jump in IPC/C combined with a decent jump in clock speed potential. My PD based APU doesn't hit i5-2400 class cpu performance until it is clocked into the 5ghz range. (not kidding, I've tested it). A $85 760K that can clock to "locked-i5-sandy" performance territory can be a great value for a performance tuning enthusiast on a budget.


In 1 thread per core vs 1 thread per module workloads:
Sandy: ~1.4X higher IPC/C than PD
Ivy: ~1.5X higher IPC/C than PD
Haswell: ~1.7X higher IPC/C than PD

In 2 thread per module vs 2 thread per non-hyper-threaded core:
PD: 1.3X higher IPC/M than Sandy-Core:
PD: 1.2X higher IPC/M than Ivy-Core:
PD: 1.05X higher IPC/M than Haswell-Core:

In 2 thread per hyper-threaded core vs 2 thread per module:
Sandy HT-Core = PD Module
Ivy HT-Core: 1.1X higher IPC/C than PD module
Haswell HT-Core: 1.25X higher IPC/C than PD module



When comparing the FX-8350 to an i5, MOST real-time workloads fall into "group 1" above where performance is compared 1 thread per core vs 1 thread per module.

Group "3" conditions above, come into play in real-time workloads when comparing CPUs with less overall parallelism, like the i3's vs 760K (HT core vs PD module, on a smaller scale).

Even the older Sandy bridge i5-2500K has no trouble out-performing an FX-8350 in 99% of CPU intensive games due to the way things stack up in that 1 thread per core vs 1 thread per module scenario. We have to go all the way back to Core2 Era before we can find intel chips that are consistently beaten in these workloads by Piledriver. (Nahalem performs on par with PD in these conditions)

Of most interesting note to me: With the expansion of execution resources in haswell, the non-hyper-threaded Haswell cores are now nearly matching the IPC/C of an entire PD module. (1 core = 1 module). Not surprising when we look at the architecture on paper. The Haswell core has as many ALUs (Arithmetic Logic Units) as an entire PD module.


I believe that what "qualifies" a CPU as a "great" gaming CPU, is great performance in real-time workloads in its price class. I'm sorry, the FX-8350's performance in real-time workloads is not competitive with Intel i5-K in that ~$200 CPU price class. The extra $20-40 price tag on the i5 is totally washed out when we compare actual cost-to-implement (cooling and PSU sizing), especially now with budget B85 boards getting BIOS updates that allow them to overclock haswell K chips. The value leading 970 chipset options are now having to compete with $80-100 B85 boards in the "budget tuner" class.

Drop down to the FX-6300 or 760K; in those price classes, AMD has very competitive performance in real-time workloads. The 760K and FX-6300 are "great" gaming chips in their price class because they don't have to compete with unlocked intel parts (760K + HSF competes with i3, FX-6300+HSF competes with locked i5s). The FX-8350 is out-maneuvered by similar-cost-to-implement Intel parts far too often to take seriously anymore (E3-1230V3 and i5-4670K). It's a novelty for AMD enthusiasts more than it is a serious contender in today's CPU landscape. If they want my stamp of approval on 8 core AMD CPUs as "gaming" chips the price has to drop to $140, $160, $180 and $200 for the 8320, 8350, 9370, and 9590 respectively. Only then can these chips compete against Intel in gaming performance value.
Ah, alright, and I do agree that the FX 8 Core Processors should be priced a little less.
+Repped.
post #32 of 402
Quote:
Originally Posted by Quesoblanco View Post

Yes you will. Games that are multithreaded.
Aren't most new games multi-threaded?
Jay
(13 items)
 
  
CPUMotherboardGraphicsRAM
FX 8350 4.4Ghz at 1.36v GA 78LMT USB3 Rev 5.0 Sapphire R9 285 OC 16GB Kingston Savage Ram 
Hard DriveHard DriveCoolingOS
1TB Western Digital 500GB Seagate Barracuda Noctua NH-D15S Windows 8.1 
MonitorPowerMouseMouse Pad
Asus MG279Q Fressync OCZ ZS 650 Watt PSU Dragonwar ELE-G9 Dragonwar 
  hide details  
Reply
Jay
(13 items)
 
  
CPUMotherboardGraphicsRAM
FX 8350 4.4Ghz at 1.36v GA 78LMT USB3 Rev 5.0 Sapphire R9 285 OC 16GB Kingston Savage Ram 
Hard DriveHard DriveCoolingOS
1TB Western Digital 500GB Seagate Barracuda Noctua NH-D15S Windows 8.1 
MonitorPowerMouseMouse Pad
Asus MG279Q Fressync OCZ ZS 650 Watt PSU Dragonwar ELE-G9 Dragonwar 
  hide details  
Reply
post #33 of 402
Yes...but I'm not aware of many that can peg more than 4, or 6 tops.
RYZEN
(14 items)
 
OFFICE PC
(12 items)
 
20TB_HOMENAS
(9 items)
 
CPUMotherboardGraphicsRAM
AMD Ryzen 7 1700X ASRock Fatal1ty X370 Professional Gaming Crossfire XFX Core Radeon R9 390X 8GB D5  Corsair Vengeance LPX 64GB DDR4 CMK64GX4M4A2400C16 
Hard DriveOptical DriveCoolingOS
A bunch of swappable SSDs LG BD-ROM Thermaltake CL-P039-AL12BL-A Contac Silent Windows 7 for Work, Windows 10 for Games 
MonitorKeyboardPowerCase
Korean 32" 4K 324K Generic mechanical rainbow LED lit eVGA 850 G2 AZZA Helios 910 CSAZ-910 
MouseAudio
Logitech G500 T1 24BIT Tube USB DAC 
CPUMotherboardGraphicsRAM
AMD Athlon II X2 255 ASRock 970DE3/U3S3 Radeon X300 Kingston KHX1600C9D3/4GX 
Hard DriveHard DriveHard DriveHard Drive
8x Samsung HD204UI 7x Seagate ST2000DM001 Transcend TS16GSSD25-S 2x WD WD40EZRX 
Case
AZZA Helios 
  hide details  
Reply
RYZEN
(14 items)
 
OFFICE PC
(12 items)
 
20TB_HOMENAS
(9 items)
 
CPUMotherboardGraphicsRAM
AMD Ryzen 7 1700X ASRock Fatal1ty X370 Professional Gaming Crossfire XFX Core Radeon R9 390X 8GB D5  Corsair Vengeance LPX 64GB DDR4 CMK64GX4M4A2400C16 
Hard DriveOptical DriveCoolingOS
A bunch of swappable SSDs LG BD-ROM Thermaltake CL-P039-AL12BL-A Contac Silent Windows 7 for Work, Windows 10 for Games 
MonitorKeyboardPowerCase
Korean 32" 4K 324K Generic mechanical rainbow LED lit eVGA 850 G2 AZZA Helios 910 CSAZ-910 
MouseAudio
Logitech G500 T1 24BIT Tube USB DAC 
CPUMotherboardGraphicsRAM
AMD Athlon II X2 255 ASRock 970DE3/U3S3 Radeon X300 Kingston KHX1600C9D3/4GX 
Hard DriveHard DriveHard DriveHard Drive
8x Samsung HD204UI 7x Seagate ST2000DM001 Transcend TS16GSSD25-S 2x WD WD40EZRX 
Case
AZZA Helios 
  hide details  
Reply
post #34 of 402
Quote:
Originally Posted by FastMHz View Post

Yes...but I'm not aware of many that can peg more than 4, or 6 tops.
Ohh!! Damn that sucks.
Jay
(13 items)
 
  
CPUMotherboardGraphicsRAM
FX 8350 4.4Ghz at 1.36v GA 78LMT USB3 Rev 5.0 Sapphire R9 285 OC 16GB Kingston Savage Ram 
Hard DriveHard DriveCoolingOS
1TB Western Digital 500GB Seagate Barracuda Noctua NH-D15S Windows 8.1 
MonitorPowerMouseMouse Pad
Asus MG279Q Fressync OCZ ZS 650 Watt PSU Dragonwar ELE-G9 Dragonwar 
  hide details  
Reply
Jay
(13 items)
 
  
CPUMotherboardGraphicsRAM
FX 8350 4.4Ghz at 1.36v GA 78LMT USB3 Rev 5.0 Sapphire R9 285 OC 16GB Kingston Savage Ram 
Hard DriveHard DriveCoolingOS
1TB Western Digital 500GB Seagate Barracuda Noctua NH-D15S Windows 8.1 
MonitorPowerMouseMouse Pad
Asus MG279Q Fressync OCZ ZS 650 Watt PSU Dragonwar ELE-G9 Dragonwar 
  hide details  
Reply
post #35 of 402
Quote:
Originally Posted by jason387 View Post

Will I notice any difference in games going from a FX 6300 at 4.4Ghz to a FX 8350 at 4.4Ghz?

10-15% better performance in real-time workloads that can scale into parallelism that far wink.gif

Overclock your FX-6300 to 4.9ghz and it will actually perform better in most games than an FX-8350@4.4ghz, and about the same in those few that are highly parallel...

The problem comes down to striking a balance between work that needs to be done, and all the new work that needs to be done to support spreading out the work to more cores. We just don't get perfect performance scaling with more and more cores in these sorts of workloads.
     
CPUMotherboardGraphicsRAM
FX-8350 990X EVO R2.0 Sparkle GTX460 768MB ballistix tactical 2 x 8GB 
Hard DriveHard DriveOptical DriveCooling
Toshiiba THNSNH 256GB Enterprise RE3 1TB Asus BD combo drive Artic A30 
OSMonitorMonitorMonitor
Manjaro Linux Samsung 21.5" LCD E2009WFP E2009WFP 
PowerCase
Seasonic G 550W Modular Fractal Design Core 3500 
CPUMotherboardGraphicsRAM
FX-6300, 4.7 GHZ@1.43V GA-970A-UD3P GTX 460 768MB Mixed DIMMs. 2x4GB + 2x8GB @ 1600-8-8-8 
Hard DriveHard DriveHard DriveHard Drive
Toshiba THNSNH 19nm 256GB 1TB Spinpoint F3 WD RE3 1TB WD RE3 1TB 
Optical DriveCoolingOSOS
yes CM Seidon 120V SolydK OpenSuse 13.1 
OSOSMonitorMonitor
Linux Mint 9-32 bit // Linux Mint 17-64 bit  Manjaro Xfce Samsung 21.5" HannsG 21.5" sideways! 
KeyboardPowerCaseMouse
Sticky ATNG Rosewill Green 630W NZXT Gamma Basic Microsoft corded 
CPUMotherboardGraphicsRAM
A10-6800k 4.8GHZ @ 1.375V, 1.2GHZ iGPU Gigabyte GA-F2A85XN-WIFI HD8670D Ripjaws 
Hard DriveOptical DriveCoolingOS
Seagate ST1000DM003 Asus BC-12B1ST/BLK/B/AS Zalman CNPS5X Linux Mint 15 
MonitorKeyboardPowerCase
LG IPS224V-PN Logitec K360 FSP 400W Aurum S 80+ gold Prodigy 
Mouse
logitec M235 
  hide details  
Reply
     
CPUMotherboardGraphicsRAM
FX-8350 990X EVO R2.0 Sparkle GTX460 768MB ballistix tactical 2 x 8GB 
Hard DriveHard DriveOptical DriveCooling
Toshiiba THNSNH 256GB Enterprise RE3 1TB Asus BD combo drive Artic A30 
OSMonitorMonitorMonitor
Manjaro Linux Samsung 21.5" LCD E2009WFP E2009WFP 
PowerCase
Seasonic G 550W Modular Fractal Design Core 3500 
CPUMotherboardGraphicsRAM
FX-6300, 4.7 GHZ@1.43V GA-970A-UD3P GTX 460 768MB Mixed DIMMs. 2x4GB + 2x8GB @ 1600-8-8-8 
Hard DriveHard DriveHard DriveHard Drive
Toshiba THNSNH 19nm 256GB 1TB Spinpoint F3 WD RE3 1TB WD RE3 1TB 
Optical DriveCoolingOSOS
yes CM Seidon 120V SolydK OpenSuse 13.1 
OSOSMonitorMonitor
Linux Mint 9-32 bit // Linux Mint 17-64 bit  Manjaro Xfce Samsung 21.5" HannsG 21.5" sideways! 
KeyboardPowerCaseMouse
Sticky ATNG Rosewill Green 630W NZXT Gamma Basic Microsoft corded 
CPUMotherboardGraphicsRAM
A10-6800k 4.8GHZ @ 1.375V, 1.2GHZ iGPU Gigabyte GA-F2A85XN-WIFI HD8670D Ripjaws 
Hard DriveOptical DriveCoolingOS
Seagate ST1000DM003 Asus BC-12B1ST/BLK/B/AS Zalman CNPS5X Linux Mint 15 
MonitorKeyboardPowerCase
LG IPS224V-PN Logitec K360 FSP 400W Aurum S 80+ gold Prodigy 
Mouse
logitec M235 
  hide details  
Reply
post #36 of 402
Quote:
Originally Posted by mdocod View Post

10-15% better performance in real-time workloads that can scale into parallelism that far wink.gif

Overclock your FX-6300 to 4.9ghz and it will actually perform better in most games than an FX-8350@4.4ghz, and about the same in those few that are highly parallel...

The problem comes down to striking a balance between work that needs to be done, and all the new work that needs to be done to support spreading out the work to more cores. We just don't get perfect performance scaling with more and more cores in these sorts of workloads.
What you said is absolutely true. This is why Intel is ahead. But I do think that more games and software should be geared to making use of more cores, not just cause you have the extra cores but because the game or software needs it to perform faster. I think that making use of more cores is the way to go with ensuring max GPU utilization. CPU's have always been important for gaming and why should a person buy a CPU and then see 40% usage or less in games. Why not make use of the CPU till FPS goes so high that it doesn't matter anymore. If the GPU works its butt off so should the CPU. Games should be geared towards this end. But sadly for game developers its either too much work, too difficult or too lazy and not wanting progress.
Jay
(13 items)
 
  
CPUMotherboardGraphicsRAM
FX 8350 4.4Ghz at 1.36v GA 78LMT USB3 Rev 5.0 Sapphire R9 285 OC 16GB Kingston Savage Ram 
Hard DriveHard DriveCoolingOS
1TB Western Digital 500GB Seagate Barracuda Noctua NH-D15S Windows 8.1 
MonitorPowerMouseMouse Pad
Asus MG279Q Fressync OCZ ZS 650 Watt PSU Dragonwar ELE-G9 Dragonwar 
  hide details  
Reply
Jay
(13 items)
 
  
CPUMotherboardGraphicsRAM
FX 8350 4.4Ghz at 1.36v GA 78LMT USB3 Rev 5.0 Sapphire R9 285 OC 16GB Kingston Savage Ram 
Hard DriveHard DriveCoolingOS
1TB Western Digital 500GB Seagate Barracuda Noctua NH-D15S Windows 8.1 
MonitorPowerMouseMouse Pad
Asus MG279Q Fressync OCZ ZS 650 Watt PSU Dragonwar ELE-G9 Dragonwar 
  hide details  
Reply
post #37 of 402
Quote:
Originally Posted by jason387 View Post

What you said is absolutely true. This is why Intel is ahead. But I do think that more games and software should be geared to making use of more cores, not just cause you have the extra cores but because the game or software needs it to perform faster. I think that making use of more cores is the way to go with ensuring max GPU utilization. CPU's have always been important for gaming and why should a person buy a CPU and then see 40% usage or less in games. Why not make use of the CPU till FPS goes so high that it doesn't matter anymore. If the GPU works its butt off so should the CPU. Games should be geared towards this end. But sadly for game developers its either too much work, too difficult or too lazy and not wanting progress.


That's one of the miss conceptions people have. If the GPU is running at a steady 100% then it wont make a lick of differences if the CPU utilization is 20% or 90%. In instances like that you are getting as much FPS as you are ever going to get with that GPU.

A fast GPU can compensate for a slow CPU just as a fast CPU can't compensate for a slow GPU. They do different things and their loads scale of different things. That's why you can have a game that looks like turd and will rape your CPU and a stunning looking game that uses very little CPU time.

The issue is when a threads "overloads". To get 60 FPS everything needs to be calculated in less then 16.7ms so if thread 1 on core 0 takes 19.0ms to complete then the game stalls for that amount of time (resulting in reduced FPS). Its not really about just multi-threading. Most semi modern games spawn more then 8 threads. The real problem is in thread balancing, that is to say to be able to off load enough of operations form the main thread on to other cores so that single core stalls don't happen. The reason this is easier said the done is that when you split operations over multiple threads you have to syncing the shared resources and that reduces the speed gains and so splitting something over several threads can actually reduced performance if implemented in unsuitable situations.

So yes its very much about difficulty and by that extension time and money. If it was easy we would have had perfectly core scaling games ages ago.
Edited by Bit_reaper - 4/21/14 at 7:59pm
    
CPUMotherboardGraphicsRAM
i7 920 OC 4.0 GHz 1.35v HT on Asus P6T 1366 SLI Gigabyte GTX 970 OCZ 12GB DDR3 GOLD/Platinum mix 
Hard DriveHard DriveHard DriveHard Drive
Samsung Spinpoint F1 1TB Western digital 300GB Western digital Caviar Blue 1TB Samsung 840 EVO 250Gb 
Optical DriveCoolingOSMonitor
generic LG DVD WC'd , Supreme LT, NexXxoS Xtreme III 360, Phob... Windows 7 x64 Samsung 27" LED S27A550B 
MonitorMonitorKeyboardPower
ACER 23" x233H Dell U3415W Logitech G11 Chieftec 850W 
CaseMouseMouse PadAudio
Phanteks enthoo luxe Razer deathAdder respawn steelseries Qck Yamaha HTR-6130 AV Receiver 
AudioAudio
Yamaha NS-50B floor tower speakers miditech Audiolink II stereo sound card 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
i7 920 OC 4.0 GHz 1.35v HT on Asus P6T 1366 SLI Gigabyte GTX 970 OCZ 12GB DDR3 GOLD/Platinum mix 
Hard DriveHard DriveHard DriveHard Drive
Samsung Spinpoint F1 1TB Western digital 300GB Western digital Caviar Blue 1TB Samsung 840 EVO 250Gb 
Optical DriveCoolingOSMonitor
generic LG DVD WC'd , Supreme LT, NexXxoS Xtreme III 360, Phob... Windows 7 x64 Samsung 27" LED S27A550B 
MonitorMonitorKeyboardPower
ACER 23" x233H Dell U3415W Logitech G11 Chieftec 850W 
CaseMouseMouse PadAudio
Phanteks enthoo luxe Razer deathAdder respawn steelseries Qck Yamaha HTR-6130 AV Receiver 
AudioAudio
Yamaha NS-50B floor tower speakers miditech Audiolink II stereo sound card 
  hide details  
Reply
post #38 of 402
Quote:
Originally Posted by Bit_reaper View Post

That's one of the miss conceptions people have. If the GPU is running at a steady 100% then it wont make a lick of differences if the CPU utilization is 20% or 90%. In instances like that you are getting as much FPS as you are ever going to get with that GPU.

A fast GPU can compensate for a slow CPU just as a fast CPU can't compensate for a slow GPU. They do different things and their loads scale of different things. That's why you can have a game that looks like turd and will rape your CPU and a stunning looking game that uses very little CPU time.

The issue is when a threads "overloads". To get 60 FPS everything needs to be calculated in less then 16.7ms so if thread 1 on core 0 takes 19.0ms to complete then the game stalls for that amount of time (resulting in reduced FPS). Its not really about just multi-threading. Most semi modern games spawn more then 8 threads. The real problem is in thread balancing, that is to say to be able to off load enough of operations form the main thread on to other cores so that single core stalls don't happen. The reason this is easier said the done is that when you split operations over multiple threads you have to syncing the shared resources and that reduces the speed gains and so splitting something over several threads can actually reduced performance if implemented in unsuitable situations.

So yes its very much about difficulty and by that extension time and money. If it was easy we would have had perfectly core scaling games ages ago.
Then how come different CPU's get different FPS( higher or lower) even with the same GPU? The CPU does matter and should matter smile.gif
Jay
(13 items)
 
  
CPUMotherboardGraphicsRAM
FX 8350 4.4Ghz at 1.36v GA 78LMT USB3 Rev 5.0 Sapphire R9 285 OC 16GB Kingston Savage Ram 
Hard DriveHard DriveCoolingOS
1TB Western Digital 500GB Seagate Barracuda Noctua NH-D15S Windows 8.1 
MonitorPowerMouseMouse Pad
Asus MG279Q Fressync OCZ ZS 650 Watt PSU Dragonwar ELE-G9 Dragonwar 
  hide details  
Reply
Jay
(13 items)
 
  
CPUMotherboardGraphicsRAM
FX 8350 4.4Ghz at 1.36v GA 78LMT USB3 Rev 5.0 Sapphire R9 285 OC 16GB Kingston Savage Ram 
Hard DriveHard DriveCoolingOS
1TB Western Digital 500GB Seagate Barracuda Noctua NH-D15S Windows 8.1 
MonitorPowerMouseMouse Pad
Asus MG279Q Fressync OCZ ZS 650 Watt PSU Dragonwar ELE-G9 Dragonwar 
  hide details  
Reply
post #39 of 402
Quote:
Originally Posted by jason387 View Post

Then how come different CPU's get different FPS( higher or lower) even with the same GPU? The CPU does matter and should matter smile.gif

Well that's because thous games have stalls on even really fast CPU's that is to say you don't see the GPU running at 100% it dips down to 99% or 98% every now and again. You might not pay much attention if you have a few dropped frames here and there but they do show up in the bench mark numbers. In some games you see deeper dips. It all depends on the game and the system spec.

Is about the bottleneck. Either you are limited by the CPU or the GPU. If you run a game at 1080p and the CPU can out put 80FPS and the GPU can only do 50 you get 50. If your GPU can do 60 but your CPU can only do 40 you get 40.

As GPU load scales largely of resolution dropping down 900p will change the GPU, CPU balance. That's why you often see CPU scaling bench marks include low resolution test.






As an game runs the GPU and CPU load constantly fluctuates independently form another. Lets there's game where there is a big explosion with lots of physics objects that cause the physics thread to max out a core well then the FPS will dips as will the GPU usage.

That's really the thing most people seem to have the hardest to grasp. That the CPU/GPU load ratio fluctuates. You can be limited by both your CPU and GPU in the same game depending on whats happening.

Another example would be a multi-player shooter when lots of players bunch up and there is lots of bullet trajectories and hit-boxes to calculate again causing that thread to max out a core and stall the game. Same result.

So yes CPU most definitely matters but making demands that the CPU should run constantly at 100% and anything less is devs been lazy or unskilled well that's just silly as it goes against how games actually work on PC's.

If what you are referring to is the little bit of extra top end FPS you sometimes see when going form an moderately fast CPU to an really fast CPU like say from an stock 3570k to an OC'd 3930K and get and increase form 130FPS to 133FPS (I'm to totally sure where it comes form but I suspect its form the flip queue --> driver --> GPU time or it could just be related to windows scheduling delays). Yes its is there and it's measurable but it rarely amounts to more then a few stray FPS that run in the 1-3% margins.

Just look up some CPU scaling graph and you should star to see what I'm talking about.
Edited by Bit_reaper - 4/22/14 at 8:26am
    
CPUMotherboardGraphicsRAM
i7 920 OC 4.0 GHz 1.35v HT on Asus P6T 1366 SLI Gigabyte GTX 970 OCZ 12GB DDR3 GOLD/Platinum mix 
Hard DriveHard DriveHard DriveHard Drive
Samsung Spinpoint F1 1TB Western digital 300GB Western digital Caviar Blue 1TB Samsung 840 EVO 250Gb 
Optical DriveCoolingOSMonitor
generic LG DVD WC'd , Supreme LT, NexXxoS Xtreme III 360, Phob... Windows 7 x64 Samsung 27" LED S27A550B 
MonitorMonitorKeyboardPower
ACER 23" x233H Dell U3415W Logitech G11 Chieftec 850W 
CaseMouseMouse PadAudio
Phanteks enthoo luxe Razer deathAdder respawn steelseries Qck Yamaha HTR-6130 AV Receiver 
AudioAudio
Yamaha NS-50B floor tower speakers miditech Audiolink II stereo sound card 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
i7 920 OC 4.0 GHz 1.35v HT on Asus P6T 1366 SLI Gigabyte GTX 970 OCZ 12GB DDR3 GOLD/Platinum mix 
Hard DriveHard DriveHard DriveHard Drive
Samsung Spinpoint F1 1TB Western digital 300GB Western digital Caviar Blue 1TB Samsung 840 EVO 250Gb 
Optical DriveCoolingOSMonitor
generic LG DVD WC'd , Supreme LT, NexXxoS Xtreme III 360, Phob... Windows 7 x64 Samsung 27" LED S27A550B 
MonitorMonitorKeyboardPower
ACER 23" x233H Dell U3415W Logitech G11 Chieftec 850W 
CaseMouseMouse PadAudio
Phanteks enthoo luxe Razer deathAdder respawn steelseries Qck Yamaha HTR-6130 AV Receiver 
AudioAudio
Yamaha NS-50B floor tower speakers miditech Audiolink II stereo sound card 
  hide details  
Reply
post #40 of 402
Quote:
Originally Posted by Bit_reaper View Post

Well that's because thous games have stalls on even really fast CPU's that is to say you don't see the GPU running at 100% it dips down to 99% or 98% every now and again. You might not pay much attention if you have a few dropped frames here and there but they do sow up in the bench mark numbers. In some games you see deeper dips. It all depends on the game and the system spec.

Is about the bottleneck. Either you are limited by the CPU or the GPU. If you run a game at 1080p and the CPU can out put 80FPS and the GPU can only do 50 you get 50. If your GPU can do 60 but your CPU can only do 40 you get 40.

As GPU load scales largely of resolution dropping down 900p will change the GPU, CPU balance. That's why you often see CPU scaling bench marks include low resolution test.






As an game runs the GPU and CPU load constantly fluctuates independently form another. Lets there's game where there is a big explosion with lots of physics objects that cause the physics thread to max out a core well then the FPS will dips as will the GPU usage.

That's really the thing most people seem to have the hardest to grasp. That the CPU/GPU load ratio fluctuates. You can be limited by both your CPU and GPU in the same game depending on whats happening.

Another example would be a multi-player shooter when lots of players bunch up and there is lots of bullet trajectories and hit-boxes to calculate again causing that thread to max out a core and stall the game. Same result.

So yes CPU most definitely matters but making demands that the CPU should run constantly at 100% and anything less is devs been lazy or unskilled well that's just silly as it goes against how games actually work on PC's.

If what you are referring to is the little bit of extra top end FPS you sometimes see when going form an moderately fast CPU to an really fast CPU like say from an stock 3570k to an OC'd 3930K and get and increase form 130FPS to 133FPS (I'm to totally sure where it comes form but I suspect its form the flip queue --> driver --> GPU time or it could just be related to windows scheduling delays). Yes its is there and it's measurable but it rarely amounts to more then a few stray FPS that run in the 1-3% margins.

Just look up some CPU scaling graph and you should star to see what I'm talking about.

Nice read. Looks like the CPU only matters at lower resolutions. My native resolution is 1366x768.
Jay
(13 items)
 
  
CPUMotherboardGraphicsRAM
FX 8350 4.4Ghz at 1.36v GA 78LMT USB3 Rev 5.0 Sapphire R9 285 OC 16GB Kingston Savage Ram 
Hard DriveHard DriveCoolingOS
1TB Western Digital 500GB Seagate Barracuda Noctua NH-D15S Windows 8.1 
MonitorPowerMouseMouse Pad
Asus MG279Q Fressync OCZ ZS 650 Watt PSU Dragonwar ELE-G9 Dragonwar 
  hide details  
Reply
Jay
(13 items)
 
  
CPUMotherboardGraphicsRAM
FX 8350 4.4Ghz at 1.36v GA 78LMT USB3 Rev 5.0 Sapphire R9 285 OC 16GB Kingston Savage Ram 
Hard DriveHard DriveCoolingOS
1TB Western Digital 500GB Seagate Barracuda Noctua NH-D15S Windows 8.1 
MonitorPowerMouseMouse Pad
Asus MG279Q Fressync OCZ ZS 650 Watt PSU Dragonwar ELE-G9 Dragonwar 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: AMD CPUs
Overclock.net › Forums › AMD › AMD CPUs › Questions about the FX 8350?