HT off for BF4 = amazing smooth performance!

Im betting win7 issue not hyperthreading issue.
Click to expand...

Possibly, or something else. I finally got a chance to graph and I do not have any where close to that activity that Shiz has with HT enabled and 8.1 OS.

BTW Shiz, you should enabled kernel time view in Task Manager. I never understand as to why people do not enable this!

Edit: I noticed that your C-states and C1E are disabled. Along with Visualization, guess you do not use any/all of them (mainly due to overclocking). However, that may be the issue are the C-States causing latency when power gating the cores-especially with HT. This could be causing errant results.
 
Last edited by a moderator:
Does core parking (un parking) have a similar effect to disabling hyperthreading?
Click to expand...

Core parking is a bullshit theory. Basically "parking" is code that Intel supplied MS for giving execution to physical cores rather than any core in Windows 7 during low execution. Once ACPI detects activity necessary the core is "unparked" and execution is given to physical and logical core as execution scales. So, disabling can give a bit of "better performance" because all execution is handled with all phys/log available, no scaling involved at that clock cycle. Parking is not an issue, if the game is loading or even being played all core are being used.

I noticed that Windows 8 is like Vista, not like 7. All cores are used; nothing seems to be parked.
 
§·H·ï·Z·N·ï·L·T·ï;1040332051 said:

Vid on results...
http://youtu.be/yMGmY6qQWQY?t=57s&hd=1

@0:00-1:00 bonus footage of my dog and turtle :D

@3:00 HT on gameplay, notice huge yellow spikes, when I shoot I get spikes.

@3:50 CPU/GPU usage results, notice the GPU usage has lots of dips, CPU usage is low due to being spread out across all 8 threads.

@7:30 HT off gameplay, notice how there's hardly any spikes even when shooting. Overall 5-10 more FPS on average and before when looking at the sky max fps was fluctuating like crazy from 80-100fps with very large spikes on the graph, now it's @ 200 and the absolutely no spikes at all with HT off.

@8:40 CPU/GPU usage results. notice how the GPU usage pegged right up at 100% no dips at all, CPU usage is higher, but more level.
Click to expand...
Unfortunately, I was NOT able to replicate the spikes like you have in the graph, at least not while firing the gun or regular run and gunning. During run and gun I was getting what looked like square sine waves. At most it fluctuated 5-10% cpu use during that time, and my FPS remained a constant 40-50 fps. Of interesting note, on a siege of shanghai map after the building collapsed and there was a ton of debris, I would get spikes randomly doing nothing while standing on the bridge near the building. Sometimes a piece of floating debris came by and there would be a spike. So far, I have yet to finish that level without it crashing during the building crash...so something IS AMISS somewhere, just not sure what.

As a test, I tried both the 32bit (bf4_x86.exe) and 64bit (bf4.exe) versions of the game.

In fact, when I noticed my hard drive thrashing again, I got pissed, shrunk it from the system managed 16GB to a fixed size of 4GB, rebooted, defragged, all my problems went away. I also manually updated Punkbuster. I hadn't even disabled HT. I'm betting my pagefile was heavily fragmented. My HDD is in excellent health according to smart tests.

That being said, I will keep testing. Knowing my luck, it will return. One thing I have noticed between our setups is I am running Win7 64bit with 16GB of RAM and you are running Win8.1. Also any possibility the lack of RAM is causing it? Yesterday I was using over 7.5GB while playing. I am also at 4.3ghz cpu OC, where as you are at 4.9. I also do not have any C states disabled. Is it possible your CPU is near maxed and the HT resources aren't getting enough juice? Maybe try at 4.5 or 4.3 like mine and see what happens.

Sorry.
 
Last edited:
Shikami said:
Core parking is a bullshit theory. Basically "parking" is code that Intel supplied MS for giving execution to physical cores rather than any core in Windows 7 during low execution. Once ACPI detects activity necessary the core is "unparked" and execution is given to physical and logical core as execution scales. So, disabling can give a bit of "better performance" because all execution is handled with all phys/log available, no scaling involved at that clock cycle. Parking is not an issue, if the game is loading or even being played all core are being used.

I noticed that Windows 8 is like Vista, not like 7. All cores are used; nothing seems to be parked.
Click to expand...

Agreed.
 
§·H·ï·Z·N·ï·L·T·ï;1040332051 said:

Vid on results...
http://youtu.be/yMGmY6qQWQY?t=57s&hd=1

@0:00-1:00 bonus footage of my dog and turtle :D

@3:00 HT on gameplay, notice huge yellow spikes, when I shoot I get spikes.

@3:50 CPU/GPU usage results, notice the GPU usage has lots of dips, CPU usage is low due to being spread out across all 8 threads.

@7:30 HT off gameplay, notice how there's hardly any spikes even when shooting. Overall 5-10 more FPS on average and before when looking at the sky max fps was fluctuating like crazy from 80-100fps with very large spikes on the graph, now it's @ 200 and the absolutely no spikes at all with HT off.

@8:40 CPU/GPU usage results. notice how the GPU usage pegged right up at 100% no dips at all, CPU usage is higher, but more level.
Click to expand...



I only have one thing to say, that song is fucking awesome, thanks for turning me on to acid kings bro :cool:
 
well.... I turned off HT and my 780s were being utilized more. Hope HT gets fixed in this game.
 
I'll give this a shot tonight on W8.1 (4770K 4.6, 2 GTX670FTWs) and report my findings....
 
Using an Ivy Bridge here and I'm only seeing 25% CPU usage across all eight threads. It has been running great for me, no complaints here!
 
Can you not over clock? You can run the multicore enhancement that all EFI have to enable. Just have your system in a normal-somewhat-manner and then test.
 
I disabled hyper threading when battlefield 3 came out and never re-enabled it. Just wondering if core parking still needs to be disabled for better performance (they're probably disabled because I have hyper threading off. But you might never know)?
 
Mr Mean said:
I disabled hyper threading when battlefield 3 came out and never re-enabled it. Just wondering if core parking still needs to be disabled for better performance (they're probably disabled because I have hyper threading off. But you might never know)?
Click to expand...

no. honestly, it's better to enable HT and just run the core parking program.


I did give bf4 a try with HT off last night(via the BIOS). I did notice a difference. I also recommend it as you have nothing to lose. It allowed me to go up to everything @ HIGH except textures = ULTRA. MSAA = off. POST AA= max. 75FPS or so, did drop and hiccup a handful of times but I think it was due to the server.

2600k @ 4.4, 6gb ddr3, SSD, 670gtx, 144hz 24" @ 1920x1080
 
Do you guys think there is a difference between HT on Sandy and Ivy and Haswell?
Also you think windows 8.1 is just utilizing HT better than w7?
I ask because my sig rig is smoother than a babies ass in BF4 with HT on and 4770k stock clocks...
I could not imagine it being smoother, in fact that's one of the first things that struck me was HOW SMOOTH BF4 is.... before the servers crash the fuck out anyway
 
So I tried disabling HT last night on my 4770K with Windows 8.1. I can't say I noticed much of a diff - it was pretty darn smooth before, with HT on. Probably ought to get more concrete numbers/validation, but it seemed the same.
 
With HT off, I'm getting 99% GPU usage on both cards (6950s), when before in the same scene (training map) with HT on I was seeing 70% per card at best. Framerate went up accordingly, and it's much more stable. Using AMD 13.11 beta 7 drivers.
 
I'm going to try this over the weekend and come back with my results. I have a 2600k and GTX670 on 64bit Win7.
 
I don't have the game, but I'm really curious about HT on different chipset platforms like Sandy Bridge, Ivy Bridge, Haswell, etc.
 
I don't have the game, but I'm really curious about HT on different chipset platforms like Sandy Bridge, Ivy Bridge, Haswell, etc.
Click to expand...

Yes, there are architectural changes for HT to IVY. Haswell, unsure of, but is either the same or has some changes.


There are other tweaks to improve Hyper-Threading (a few queues are now partitioned dynamically between two threads, rather than shared statically at 50-50) and AVX performance (more registers to help deal with memory access that cross cache lines).
Click to expand...

http://techreport.com/review/22835/review-intel-core-i7-3770k-ivy-bridge-processor/2
 
Well I'm pegged at 99 percent GPU usage and around 54 percent CPU usage in BF4. Weird that some of us have greater CPU usage while some of us have maxed GPU usage and some do not. Frostbite 3 is an odd engine.

I also tried Shadowplay and basically took a 5fps performance hit. What a great addition to the driver set.
 
I am not having any issues myself and it plays awesome on my i7-930 with HT on at stock speed and an HD7950 Boost with windows 8 with just 6Gb of ram at 1920 x 1080 with Ultra settings with Post off on 34 man server.
 
Turned my HT off i7860 at stock speeds and 7870. Plays great! Thanks you.
 
fanboy said:
I am not having any issues myself and it plays awesome on my i7-930 with HT on at stock speed and an HD7950 Boost with windows 8 with just 6Gb of ram at 1920 x 1080 with Ultra settings with Post off on 34 man server.
Click to expand...

I wanted to instantly call BS on your claim, but you didn't post any concrete FPS numbers..But having played through the BeTA and seeing the beating that some highly O/C'd CPUs were taking, there is NO WAY that you are getting anywhere near 50-60 FPS(I know you didn't claim a solid number but for an FPS play to be "awesome" you need at least 45+FPS on the low end so it isn't a laggy slide show) with your 930 @ stock speed on ULTRA...ULTRA settings eat CPU cycles for breakfast, lunch, and dinner...If you step up to a full 64 player map you will have a slide show..

Also the 930 HAS to be holding your 7950 Boost back quite a bit..Give it a quick OC to ~3.5-3.6Ghz and do a before/after FRAPS run..I bet you gain double digits...

I am not trying to insult you or attack you personally, just pointing out the performance potential you are missing with HW you are already running!:eek::eek:
 
ill have to try this. i know bf3 benefited from ht off.

but from what i heard is that bf4 make good uses of multiple threads.
 
ccityinstaller said:
I wanted to instantly call BS on your claim, but you didn't post any concrete FPS numbers..But having played through the BeTA and seeing the beating that some highly O/C'd CPUs were taking, there is NO WAY that you are getting anywhere near 50-60 FPS(I know you didn't claim a solid number but for an FPS play to be "awesome" you need at least 45+FPS on the low end so it isn't a laggy slide show) with your 930 @ stock speed on ULTRA...ULTRA settings eat CPU cycles for breakfast, lunch, and dinner...If you step up to a full 64 player map you will have a slide show..

Also the 930 HAS to be holding your 7950 Boost back quite a bit..Give it a quick OC to ~3.5-3.6Ghz and do a before/after FRAPS run..I bet you gain double digits...

I am not trying to insult you or attack you personally, just pointing out the performance potential you are missing with HW you are already running!:eek::eek:
Click to expand...

^ this, even cs go runs better with a good OC
 
Hmmm i was getting a small stutter on some maps. Figure it was drivers.....Might try turning off HT to see if that fixes it. Ill post more when I get home in the morning.
 
This thread inspired me to do some personal testing since I was thinking of turning on my HT for streaming purposes. I do want a new video card but currently I can survive with custom medium settings until Black Friday or my birthday in December.

In-Game settings: 1920x1080 fullscreen, High Textures, Medium everything else except HBAO/AA settings set to off. System specs in sig, using Catalyst 13.9. Tests were conducted on best nearby servers in US West Coast, Conquest or Obliteration, 42-64 player servers and tried to stay on the same server if I can. Avg test time was about 10-15 minutes.

Take the following test results as anecdotal: I did these without a FPS meter, but rather eyeballing it. My goal was to see if its a good gameplay experience: doesn't matter if I get high FPS - if I can tell there's input lag or something's off then its no good. Basically I'm trying to squeeze out what I can out of a 5850 with 50-60 avg FPS being the goal.

1st Test: 3.4ghz with HT off + power saving stuff off (currently my setup) = relatively silky smooth (50+), can tell if the server is terrible or not.
2nd Test: 3.4ghz with HT on + ^ = can eyeball lower FPS, weird input/sound lag (1-3 second on avg)
3rd Test: 2nd Test + C1E / Speedstep / C3-C6 turned on = same result as 2nd Test + stuttering.
4th Test: Stock 920 settings + HT on + default BIOS settings = Same as 3rd Test
5th Test: Stock 920 settings + HT off + power saving settings off = much better but I can tell my FPS was not as stable as it was at 3.4ghz.

I want to do another series of tests with actually spending time tweaking my OC properly rather than a rush job, and more stable servers. I know a 920 can be pushed to at least 3.8+Ghz but running into BSOD's after 3.6ghz with 6 RAM sticks. The 3.4ghz OC does work after using Prime95/OCCT for 12 hrs and was running BF3 fine (other games not mentioned cause they're run on a Core 2 without effort).
 
Kaiyoko-Desu said:
This thread inspired me to do some personal testing since I was thinking of turning on my HT for streaming purposes. I do want a new video card but currently I can survive with custom medium settings until Black Friday or my birthday in December.

In-Game settings: 1920x1080 fullscreen, High Textures, Medium everything else except HBAO/AA settings set to off. System specs in sig, using Catalyst 13.9. Tests were conducted on best nearby servers in US West Coast, Conquest or Obliteration, 42-64 player servers and tried to stay on the same server if I can. Avg test time was about 10-15 minutes.

Take the following test results as anecdotal: I did these without a FPS meter, but rather eyeballing it. My goal was to see if its a good gameplay experience: doesn't matter if I get high FPS - if I can tell there's input lag or something's off then its no good. Basically I'm trying to squeeze out what I can out of a 5850 with 50-60 avg FPS being the goal.

1st Test: 3.4ghz with HT off + power saving stuff off (currently my setup) = relatively silky smooth (50+), can tell if the server is terrible or not.
2nd Test: 3.4ghz with HT on + ^ = can eyeball lower FPS, weird input/sound lag (1-3 second on avg)
3rd Test: 2nd Test + C1E / Speedstep / C3-C6 turned on = same result as 2nd Test + stuttering.
4th Test: Stock 920 settings + HT on + default BIOS settings = Same as 3rd Test
5th Test: Stock 920 settings + HT off + power saving settings off = much better but I can tell my FPS was not as stable as it was at 3.4ghz.

I want to do another series of tests with actually spending time tweaking my OC properly rather than a rush job, and more stable servers. I know a 920 can be pushed to at least 3.8+Ghz but running into BSOD's after 3.6ghz with 6 RAM sticks. The 3.4ghz OC does work after using Prime95/OCCT for 12 hrs and was running BF3 fine (other games not mentioned cause they're run on a Core 2 without effort).
Click to expand...

Make sure when you are pushing the OC that you feed some voltage to the IMC, especially since you are using all 6 DIMM slots...Keep your IMC voltage within .5V of the DIMM Voltage as well, or you will cook the IMC pretty quickly..You can also adjust the IOH voltage up to ~1.15V and see if that helps as well..Make sure to disable Spread Spectrum etc...I never had any luck with leaving any power saving options on when O/C'ing on 1366, and since I run my CPU @ 100% anyway 24/7, it made no sense to me..
 
If your not running windows 8 or 8.1 then you don't have DX 11.1 running which is why windows 7 users need a heavy overclock to make up for it because windows 8 makes great use out of 8 treads..

I have upgraded to windows 8.1 and have bump the gpu to 1005/1503 and on Ultra settings I have not seen lower then 40fps with the cpu stock and HT.
 
You must log in or register to reply here.
Back
Top