Overclock.net › Forums › Video Games › PC Gaming › Bottlenecks That Don't Exist and 4770K vs 4960X for Poorly Threaded Games
New Posts  All Forums:Forum Nav:

Bottlenecks That Don't Exist and 4770K vs 4960X for Poorly Threaded Games

post #1 of 8
Thread Starter 
I've been wondering about a couple of things:

In most games you can clearly see the bottlenecking component: A GPU or CPU cores saturated. But what happens when you get something like this:



The framerate is only at 50fps, but there are no artificial caps (eg. V-Sync) and neither the GPU/s, nor any of the CPU cores are fully loaded. What is the limiting factor here?

Another example would be Crysis Warhead. Now in Crysis, the bottleneck is usually crystal clear - you see one CPU core at 100%. But in warhead I will often see the following:



Again, moderate-low usage on the GPUs and CPU cores, no V-Sync or cap, and yet the framerate sits at a miserable 45fps.

Why?



Something that may or may not be related to the Crysis case:

Haswell is supposed to have the higher IPC, so you'd expect it to provide better performance in older games that are poorly threaded and CPU bound. And this is usually verified by benchmarks that focus on such scenarios, for example:


(Source)

However, I've seen more than one Crysis benchmark like this:


(Source)

As you can see, IB-E is well ahead of Haswell in this CPU bound benchmark, despite the fact that Crysis is poorly threaded and can't possibly benefit from the 4960X's extra cores.

So what's making the difference? Cache? RAM bandwidth?

If I'm mostly interested in older games that can't use more than 4 threads, the 4770K seems like a no-brainer. But should I be thinking beyond raw IPC?
Edited by Oubadah - 9/28/13 at 3:46pm
Skylake I
(10 items)
 
Skylake II
(9 items)
 
 
CPUMotherboardRAMHard Drive
Intel Core i7-6700K 4GHz Gigabyte GA-Z170X-Designare F23a Corsair LPX 32GB CMK32GX4M4A2133C13 Samsung SSD 840 PRO 256GB 
Hard DriveCoolingOSPower
Samsung SSD 840 PRO 256GB Noctua NH-U12S (+NF-S12A PWM) Windows 7 Pro x64 SP1/Windows 10 Pro x64 Super Flower Leadex Gold 750W 
CaseAudio
Corsair Carbide Air 540 (+3x NF-S12A FLX) Asus Xonar Essence STX 
CPUMotherboardRAMHard Drive
Intel Core i7-6700K 4GHz Gigabyte GA-Z170X-Designare F23a Corsair LPX 32GB CMK32GX4M4A2133C13 Samsung SSD 840 PRO 512GB 
Hard DriveCoolingOSPower
Samsung SSD 840 PRO 512GB Noctua NH-U12S (+NF-S12A PWM) Windows 7 Pro x64 SP1/Windows 10 Pro x64 EVGA SuperNOVA G2 750W 
Case
Silverstone Fortress FT02 (+4x NF-A14 PWM) 
  hide details  
Reply
Skylake I
(10 items)
 
Skylake II
(9 items)
 
 
CPUMotherboardRAMHard Drive
Intel Core i7-6700K 4GHz Gigabyte GA-Z170X-Designare F23a Corsair LPX 32GB CMK32GX4M4A2133C13 Samsung SSD 840 PRO 256GB 
Hard DriveCoolingOSPower
Samsung SSD 840 PRO 256GB Noctua NH-U12S (+NF-S12A PWM) Windows 7 Pro x64 SP1/Windows 10 Pro x64 Super Flower Leadex Gold 750W 
CaseAudio
Corsair Carbide Air 540 (+3x NF-S12A FLX) Asus Xonar Essence STX 
CPUMotherboardRAMHard Drive
Intel Core i7-6700K 4GHz Gigabyte GA-Z170X-Designare F23a Corsair LPX 32GB CMK32GX4M4A2133C13 Samsung SSD 840 PRO 512GB 
Hard DriveCoolingOSPower
Samsung SSD 840 PRO 512GB Noctua NH-U12S (+NF-S12A PWM) Windows 7 Pro x64 SP1/Windows 10 Pro x64 EVGA SuperNOVA G2 750W 
Case
Silverstone Fortress FT02 (+4x NF-A14 PWM) 
  hide details  
Reply
post #2 of 8
Gamebyro likes cache. In older benchmarks you can see quite the difference between an i5 2500k and an i7 2600k in fallout 3.

Also you can't trust windows to report cpu usage 100% accurately. That really isn't telling you much.
My System
(14 items)
 
  
CPUMotherboardGraphicsRAM
i5 3570k Gigabyte Z77X-UD5H MSI TF3 7950 Tri-fire Samsung 30nm 4x4GB 
Hard DriveOSMonitorKeyboard
Intel 320 160GB and a few samsung platters Windows7 x64 Catleap Q270 Rosewill mechanical 
PowerCaseMouseMouse Pad
Corsair AX850 Xigmatek Elysium Razer Deathadder Puretrak talent 
Audio
Xonar Essence STX 
  hide details  
Reply
My System
(14 items)
 
  
CPUMotherboardGraphicsRAM
i5 3570k Gigabyte Z77X-UD5H MSI TF3 7950 Tri-fire Samsung 30nm 4x4GB 
Hard DriveOSMonitorKeyboard
Intel 320 160GB and a few samsung platters Windows7 x64 Catleap Q270 Rosewill mechanical 
PowerCaseMouseMouse Pad
Corsair AX850 Xigmatek Elysium Razer Deathadder Puretrak talent 
Audio
Xonar Essence STX 
  hide details  
Reply
post #3 of 8
Thread Starter 
This Hitman Absolution benchmark shows something similar to the above Crysis phenomenon:


(Source)

The reviewer seems to be suggesting cache or memory bandwidth:
Quote:
With clock differences fairly small this leaves us to give the credit to either the memory bandwidth advantage or the larger cache.

I'm almost leaning towards the 4960X, despite the fact that it would leave me with nothing left for my SSD scheme (or the new AMD card that would probably take precedence). I am also loath to purchase such a horribly outdated platform. I wish they'd at least given us a new chipset for IB-E.
Quote:
Originally Posted by BababooeyHTJ View Post

Also you can't trust windows to report cpu usage 100% accurately. That really isn't telling you much.

A few people have said this, but what is the explanation behind it. I've noticed that if you load a single thread on a dual core processor (say, in Prime95), in task manager you see both cores partially loaded instead of one of them at 100%. I'm sure it was different in XP - I seem to remember that if you loaded one core in XP, the task manager graph would show exactly that: one core 100% loaded.

Is there any way to view it 'accurately'?

EDIT: I see what you mean about Fallout 3. I wonder why Skyrim doesn't respond the same way in 4770K vs 4960X comparisons.
Edited by Oubadah - 9/28/13 at 3:37pm
Skylake I
(10 items)
 
Skylake II
(9 items)
 
 
CPUMotherboardRAMHard Drive
Intel Core i7-6700K 4GHz Gigabyte GA-Z170X-Designare F23a Corsair LPX 32GB CMK32GX4M4A2133C13 Samsung SSD 840 PRO 256GB 
Hard DriveCoolingOSPower
Samsung SSD 840 PRO 256GB Noctua NH-U12S (+NF-S12A PWM) Windows 7 Pro x64 SP1/Windows 10 Pro x64 Super Flower Leadex Gold 750W 
CaseAudio
Corsair Carbide Air 540 (+3x NF-S12A FLX) Asus Xonar Essence STX 
CPUMotherboardRAMHard Drive
Intel Core i7-6700K 4GHz Gigabyte GA-Z170X-Designare F23a Corsair LPX 32GB CMK32GX4M4A2133C13 Samsung SSD 840 PRO 512GB 
Hard DriveCoolingOSPower
Samsung SSD 840 PRO 512GB Noctua NH-U12S (+NF-S12A PWM) Windows 7 Pro x64 SP1/Windows 10 Pro x64 EVGA SuperNOVA G2 750W 
Case
Silverstone Fortress FT02 (+4x NF-A14 PWM) 
  hide details  
Reply
Skylake I
(10 items)
 
Skylake II
(9 items)
 
 
CPUMotherboardRAMHard Drive
Intel Core i7-6700K 4GHz Gigabyte GA-Z170X-Designare F23a Corsair LPX 32GB CMK32GX4M4A2133C13 Samsung SSD 840 PRO 256GB 
Hard DriveCoolingOSPower
Samsung SSD 840 PRO 256GB Noctua NH-U12S (+NF-S12A PWM) Windows 7 Pro x64 SP1/Windows 10 Pro x64 Super Flower Leadex Gold 750W 
CaseAudio
Corsair Carbide Air 540 (+3x NF-S12A FLX) Asus Xonar Essence STX 
CPUMotherboardRAMHard Drive
Intel Core i7-6700K 4GHz Gigabyte GA-Z170X-Designare F23a Corsair LPX 32GB CMK32GX4M4A2133C13 Samsung SSD 840 PRO 512GB 
Hard DriveCoolingOSPower
Samsung SSD 840 PRO 512GB Noctua NH-U12S (+NF-S12A PWM) Windows 7 Pro x64 SP1/Windows 10 Pro x64 EVGA SuperNOVA G2 750W 
Case
Silverstone Fortress FT02 (+4x NF-A14 PWM) 
  hide details  
Reply
post #4 of 8
I'm no software developer but based on my limited understanding there are many aspects of the cpu that can be "bottlenecked".

I think that you're Oblivion screenshot is a great example. That is a notoriously single threaded game. You're being held back by the cpu and look at what windows is reporting. I'm not even sure what exactly windows is reporting with that cpu usage graph.

I would be willing to bet that if you were to benchmark Skyrim with a few mods installed it would show gains from extra cache as well. Just a hunch.
My System
(14 items)
 
  
CPUMotherboardGraphicsRAM
i5 3570k Gigabyte Z77X-UD5H MSI TF3 7950 Tri-fire Samsung 30nm 4x4GB 
Hard DriveOSMonitorKeyboard
Intel 320 160GB and a few samsung platters Windows7 x64 Catleap Q270 Rosewill mechanical 
PowerCaseMouseMouse Pad
Corsair AX850 Xigmatek Elysium Razer Deathadder Puretrak talent 
Audio
Xonar Essence STX 
  hide details  
Reply
My System
(14 items)
 
  
CPUMotherboardGraphicsRAM
i5 3570k Gigabyte Z77X-UD5H MSI TF3 7950 Tri-fire Samsung 30nm 4x4GB 
Hard DriveOSMonitorKeyboard
Intel 320 160GB and a few samsung platters Windows7 x64 Catleap Q270 Rosewill mechanical 
PowerCaseMouseMouse Pad
Corsair AX850 Xigmatek Elysium Razer Deathadder Puretrak talent 
Audio
Xonar Essence STX 
  hide details  
Reply
post #5 of 8
Skyrim has already been proven to have terrible game code in some instances. I don't think you are seeing any bottleneck except the game itself lol.

Before a few patches whiterun would randomly drops fps for no reason. You needed a script extender to fix it.
    
CPUMotherboardGraphicsRAM
Intel Core i7 2600K MSI p67 GD65  MSI GTX 680 TWIN FROZR III OC G Skill 8 GB 1600 mhz CL9 
Hard DriveHard DriveOptical DriveCooling
OCZ Vertex 3 256 GB WD 750 GB Black N/A Custom Water Cooling Loop 
OSMonitorKeyboardPower
Windows 8 Pro 64 bit Samsung un46d6420 LED 3D HDTV 46 inch Razer Black Widow Ultimate Corsair HX1050 
CaseMouseMouse PadAudio
Antec p280 Corsair M60 Razer Goliathus Alpha Sound Blaster ZXR 
AudioAudio
Harmon Kardon htfs 15 5.1 Surround Speakers run... Beyerdynamic DT 770 250 Ohm 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Intel Core i7 2600K MSI p67 GD65  MSI GTX 680 TWIN FROZR III OC G Skill 8 GB 1600 mhz CL9 
Hard DriveHard DriveOptical DriveCooling
OCZ Vertex 3 256 GB WD 750 GB Black N/A Custom Water Cooling Loop 
OSMonitorKeyboardPower
Windows 8 Pro 64 bit Samsung un46d6420 LED 3D HDTV 46 inch Razer Black Widow Ultimate Corsair HX1050 
CaseMouseMouse PadAudio
Antec p280 Corsair M60 Razer Goliathus Alpha Sound Blaster ZXR 
AudioAudio
Harmon Kardon htfs 15 5.1 Surround Speakers run... Beyerdynamic DT 770 250 Ohm 
  hide details  
Reply
post #6 of 8
Quote:
Originally Posted by BababooeyHTJ View Post

Gamebyro likes cache. In older benchmarks you can see quite the difference between an i5 2500k and an i7 2600k in fallout 3.

Also you can't trust windows to report cpu usage 100% accurately. That really isn't telling you much.

comments like this drive me nuts, no offense. True or not, his point still stands. It may not be accurate with precision in mind, but im pretty sure we can all agree it isn't reporting 39% and is actually at full load ............... so your comment was kinda pointless.
post #7 of 8
Quote:
Originally Posted by klepp0906 View Post

comments like this drive me nuts, no offense. True or not, his point still stands. It may not be accurate with precision in mind, but im pretty sure we can all agree it isn't reporting 39% and is actually at full load ............... so your comment was kinda pointless.

No offense, but ... that CPU usage pattern you see in the first screen shot? Add up all the %'s and you'll get right around 200% (out of a possible 400%), right?

Now ... the fact is, Windows can be SO wrong in it's reporting that what is ACTUALLY happening at the 'bare-metal' level is that the game is actually only using 2 cores, both fully maxed, and two cores are sitting completely idle.

Yet Windows is telling your CPU monitoring program ... that set of numbers you see on the screenshot, where all cores have a decent load on them.

That's how 'wrong' Windows CPU usage reporting can be. So his comment is actually not at all 'pointless'.

The best way to test for CPU bottlenecking is through testing/benchmarks at different clock speeds. However, not even that is not 100% reliable because if your CPU cache(s) are saturated, changing the clock speed won't help much, but it's still a 'cpu bottleneck' in effect, but ... yeah. Altering clock speeds and benching is your best shot at detecting CPU BN. Always has been, always will be.

Anything else, looking at usage % graphs, etc ... is guesswork. Although, it can become more helpful in instances where you absolutely KNOW ... how many cores/threads a game is capable of running on.
Edited by brettjv - 1/14/14 at 9:32am
    
CPUMotherboardGraphicsRAM
xeon X5675 6-core @ 4.1ghz (1.29v, 20x205 +ht ) rampage iii extreme msi rx470 gaming X (the $159 budget king) 3 x 2gb corsair xms3 pc12800 (9-9-9-24-1T@1600MHz) 
Hard DriveOptical DriveCoolingOS
hynix 250gb ssd (boot), 2tb deskstar (apps),1tb... plextor px-712sa - still the best optical drive... corsair h8o v2 aio W10 home 
MonitorPowerCaseAudio
asus vw266h 25.5" (1920x1200) abs sl (enermax revolution) * single 70A rail 850w silverstone rv-03 XFi Titanium 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
xeon X5675 6-core @ 4.1ghz (1.29v, 20x205 +ht ) rampage iii extreme msi rx470 gaming X (the $159 budget king) 3 x 2gb corsair xms3 pc12800 (9-9-9-24-1T@1600MHz) 
Hard DriveOptical DriveCoolingOS
hynix 250gb ssd (boot), 2tb deskstar (apps),1tb... plextor px-712sa - still the best optical drive... corsair h8o v2 aio W10 home 
MonitorPowerCaseAudio
asus vw266h 25.5" (1920x1200) abs sl (enermax revolution) * single 70A rail 850w silverstone rv-03 XFi Titanium 
  hide details  
Reply
post #8 of 8
Thread Starter 
Quote:
Originally Posted by brettjv View Post

No offense, but ... that CPU usage pattern you see in the first screen shot? Add up all the %'s and you'll get right around 200% (out of a possible 400%), right?

Now ... the fact is, Windows can be SO wrong in it's reporting that what is ACTUALLY happening at the 'bare-metal' level is that the game is actually only using 2 cores, both fully maxed, and two cores are sitting completely idle.

Yet Windows is telling your CPU monitoring program ... that set of numbers you see on the screenshot, where all cores have a decent load on them.

That's how 'wrong' Windows CPU usage reporting can be.

How come this doesn't happen in all games though? In so many other cases you can see that one thread running at 100%.

I tried Hitman Absolution on Haswell today. It was running 60fps, 60fps, 60fps, then I turned a corner and boom! The framerate plummets. I look at the GPU and it's ~65% loaded. I looked at the taskman CPU graph and none of the cores were anywhere near 100%. So I'm wondering if this is your alleged misreporting or something to do with the potential cache/RAM bandwidth botleneck suggested in the review benchmark I posted earlier.

Conversely, when I encountered a CPU bottleneck in Far Cry 3 the other day it was immediately identifiable in the CPU graph as one core sitting blatantly at 100%. I wonder if Windows is misreading CPU usage/averaging it out over all four graphs, why didn't it exhibit the same behaviour with Far Cry 3?
Edited by Oubadah - 3/17/14 at 10:05pm
Skylake I
(10 items)
 
Skylake II
(9 items)
 
 
CPUMotherboardRAMHard Drive
Intel Core i7-6700K 4GHz Gigabyte GA-Z170X-Designare F23a Corsair LPX 32GB CMK32GX4M4A2133C13 Samsung SSD 840 PRO 256GB 
Hard DriveCoolingOSPower
Samsung SSD 840 PRO 256GB Noctua NH-U12S (+NF-S12A PWM) Windows 7 Pro x64 SP1/Windows 10 Pro x64 Super Flower Leadex Gold 750W 
CaseAudio
Corsair Carbide Air 540 (+3x NF-S12A FLX) Asus Xonar Essence STX 
CPUMotherboardRAMHard Drive
Intel Core i7-6700K 4GHz Gigabyte GA-Z170X-Designare F23a Corsair LPX 32GB CMK32GX4M4A2133C13 Samsung SSD 840 PRO 512GB 
Hard DriveCoolingOSPower
Samsung SSD 840 PRO 512GB Noctua NH-U12S (+NF-S12A PWM) Windows 7 Pro x64 SP1/Windows 10 Pro x64 EVGA SuperNOVA G2 750W 
Case
Silverstone Fortress FT02 (+4x NF-A14 PWM) 
  hide details  
Reply
Skylake I
(10 items)
 
Skylake II
(9 items)
 
 
CPUMotherboardRAMHard Drive
Intel Core i7-6700K 4GHz Gigabyte GA-Z170X-Designare F23a Corsair LPX 32GB CMK32GX4M4A2133C13 Samsung SSD 840 PRO 256GB 
Hard DriveCoolingOSPower
Samsung SSD 840 PRO 256GB Noctua NH-U12S (+NF-S12A PWM) Windows 7 Pro x64 SP1/Windows 10 Pro x64 Super Flower Leadex Gold 750W 
CaseAudio
Corsair Carbide Air 540 (+3x NF-S12A FLX) Asus Xonar Essence STX 
CPUMotherboardRAMHard Drive
Intel Core i7-6700K 4GHz Gigabyte GA-Z170X-Designare F23a Corsair LPX 32GB CMK32GX4M4A2133C13 Samsung SSD 840 PRO 512GB 
Hard DriveCoolingOSPower
Samsung SSD 840 PRO 512GB Noctua NH-U12S (+NF-S12A PWM) Windows 7 Pro x64 SP1/Windows 10 Pro x64 EVGA SuperNOVA G2 750W 
Case
Silverstone Fortress FT02 (+4x NF-A14 PWM) 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: PC Gaming
Overclock.net › Forums › Video Games › PC Gaming › Bottlenecks That Don't Exist and 4770K vs 4960X for Poorly Threaded Games