*TBG* CPU Bottlenecking in Games (BF4 too) - HT/OC & core unparking *Win8.1 added*

[wand3r3r]

An interesting read & topic!

Includes Hitman Absolution, Far Cry 3, Tomb Raider, Crysis 3, Battlefield 4.
The core unparking was interesting, I wasn't aware of that. I'd love to see windows 8.1 BF4 numbers too.

The effect of hyperthreading (HT) on gaming performance has been much-debated for many years, especially since the advent of HT in quad-cores back in Intel's Nehalem generation of CPUs introduced in 2008. Our previous article on this topic found that HT helps dual-cores so much that you shouldn't even bother with a non-HT dual-core if you intend to play modern games. The results on a quad-core were less clear, however, so we decided we'd take a second look at it, this time adding overclocking into the mix. And as you'll see, overclocking is desperately needed with the latest system-busting games, particularly Crysis 3 and Battlefield 4. We also test a few games we've looked at previously but which are still thoroughly modern - Hitman Absolution, Tomb Raider, and Far Cry 3.

http://www.techbuyersguru.com/ochtgaming1.php

 

[wand3r3r]

Has anyone seen any 6 core tests? It appears HT is useless in windows 7 and BF4 which I found surprising since I mistakenly presumed it would be even more multithreaded now that it's presumably going to be optimized for consoles too. I'm awaiting win8.1 testing.

 

[ChronoReverse]

I don't understand why their results are different from everyone else who's done testing with HT (showing either no difference or better performance).

 

[nurturedhate]

Just anecdotal evidence here regarding bf4. Friend of mine and myself have very similar systems with the difference being his 4770k vs my 4670k. At stock there seems to be a marked difference in mp performance. Once we overclock to anything beyond 4-4.1ghz the performance gap disappears.

 

[BallaTheFeared]

Only thing concerning is Crysis 3, but I'm pretty sure I got higher fps with my i5 and 7950 CF.

I don't recall dipping below 60 fps very often, though it did happen a few times not enough to reproduce his results but that was at 4.8GHz, though it seems you need 800MHz to make up the HT difference in Crysis 3.

 

[wand3r3r]

Isn't 7950 crossfire considerably faster than a 780, enough that the CPU matters less.

 

[BallaTheFeared]

That's not the way it should work though, either the cpu is fast enough or it isn't.

You can't take a slow CPU and add more GPU power to increase FPS if the cpu was already the limiting factor.

The more gpu power, and thus more fps you can push the more the cpu actually matters.

His graph shows that at 4.5GHz HT is good for about 20% more FPS. That isn't my concern though, my concern is the FPS level he's at.

 

[frozentundra123456]

The whole article is questionable IMO. FC3, Hitman, and TR look to be gpu bound. Thus no matter what you do, you will not see any effect from hyperthreading or overclocking. The author does not even qualify the data by adding this caveat.

Crysis 3 behaves reasonably, but the BF3 results dont make any sense at all. In addition they conflict seriously with the game.gpu results for BF4, which show about a 20% gain from the 4670k to 4770k. I cant tell what map they (game.gpu) use though or if it is single player or multi, since translate does not seem to be working on my comp at work.

 

[mindbomb]

turning off hyperthreading and core unparking are surpringly frequently mentioned in game forums, despite really not affecting performance in any statistically significant way.

it's like alternative medicine...but for your fps rather than your body

 

[toyota]

I want to see this done on 8.1.

and btw Hitman and even Tomb Raider have some spots that are very cpu bound. heck just run the Hitman benchmark on highest settings and no AA and you cant stay above 60 fps with a 2500k no matter what resolution used.

 

[ozzy702]

I just un-parked my cores and it made a huge difference in how BF4 played in SLI. With cores parked it was very jerky feeling despite frame rates being high.

 

[wand3r3r]

I just un-parked my cores and it made a huge difference in how BF4 played in SLI. With cores parked it was very jerky feeling despite frame rates being high.

So it's really smooth now? I wonder if it's possible to quantify what it did via FCAT or similar methodology (or FRAPS)?

 

[ozzy702]

So it's really smooth now? I wonder if it's possible to quantify what it did via FCAT or similar methodology (or FRAPS)?

It made a dramatic difference and I'm sure there is a scientific way to quantify what my 20 years of PC gaming calibrated eyes are telling me.

The game plays very smooth now (assuming the server isn't crap) whereas I was having serious stutter before hand. Frame rates in fraps are showing about the same, maybe a bit higher, but the feeling of smoothness is completely different. Of course I only run fraps up in the corner so I can't post charts or anything like that.

I have not, nor will I turn off hyperthreading. I'm using the following hardware.

i7-3770k @ 4.4ghz
Z77-Sabertooth
SLI GTX 780's
8GB 2133 GSKILL
Samsung 840 Pro

 

[Carfax83]

The reviewer should have used Windows 8/8.1, rather than that old cuss Windows 7

Windows 8 is more efficient with CPU cycles and management, so it's possible that all the strange performance issues with HT would have been completely mitigated..

 

[AtenRa]

FX8350 on Win 8,
Cores are not parked even on an empty server. First image empty, second image with 64 players.

 

[sm625]

Most of the data is within any reasonable margin of error. I found nothing definitive in that article. (Except that BF4 is certainly faster at 4.5GHz vs 3.3GHz) Seriously, you cannot glean any more information than that once you factor in acceptable margins of error on all the bar charts.

 

[wand3r3r]

What are you referring to with the within the margin of error?

HT clearly shows a significant negative effect in BF4 in win7 (no margin of error should clarify that huge drop). That's the biggest game I'm personally curious about at this time.

 

[Carfax83]

What are you referring to with the within the margin of error?

HT clearly shows a significant negative effect in BF4 in win7 (no margin of error should clarify that huge drop). That's the biggest game I'm personally curious about at this time.

NVidia drivers has an issue with Windows 7 apparently, that causes performance drops in BF4. Using Windows 8/8.1 fixes the problem.

 

[wand3r3r]

Do you have any sources about that (BF4)? I've only seen speculation etc. so far.

 

[wand3r3r]

They added an addendum (additional page) with Windows 8.1.

How does one summarize results that are clearly all over the map? In one game, Crysis 3, you simply cannot have enough CPU power, and Hyperthreading looks like the greatest thing since sliced bread, especially under Windows 7. In Battlefield 4, the much-anticipated follow-up to the extremely popular Battlefield 3, we're let down by Hyperthreading, just as we were in BF3. Windows 8.1 helps alleviate the problem, but it does not cure it. In every test we ran, both of our systems were slower with HT enabled than with it disabled.

Windows 8.1 clearly has some improvements.

There are some discrepancies to be aware of on the 8.1 testing page, such as the 4770k at 4.2GHz whereas the previous testing on windows 7 was a 3770k at 4.5GHz. It still gives the general picture anyway.

Rather than upgrade our Windows 7 system to Windows 8.1, however, we decided to simply build a second benchmarking rig, which allows us to conduct back-to-back comparisons. This machine, using an i7-4770K at 4.2GHz, is speed-matched as closely as possible to the original machine, which uses a 3770K at 4.5GHz.

http://www.thetechbuyersguru.com/ochtgaming1.php
^The article

 

[moonbogg]

NVidia drivers has an issue with Windows 7 apparently, that causes performance drops in BF4. Using Windows 8/8.1 fixes the problem.

That was my experience. Also, regarding HT not helping with BF4, I find that suspicious to say the least. Too much info already out there, including testing done here by us forums users, that HT and more threads benefits BF3 and BF4 significantly.
I suspect, due to all the conflicting info regarding core parking, windows versions etc, that something wonky is going on in an inconsistent way with BF4. Its just too early to know for sure whats going on IMO.

 

[mindbomb]

though it isn't really marketed at all, i think haswell's hyperthreading hardware may have been improved over ivybridge.

 

[JoeRambo]

turning off hyperthreading and core unparking are surpringly frequently mentioned in game forums, despite really not affecting performance in any statistically significant way.

it's like alternative medicine...but for your fps rather than your body

How about a very repeatable test in Win7 for 4770K:

1) Start BF4
2) Load test range
3) Enable perf graph in console (PerfOverlay.DrawGraph 1)
4) Wander around for some time, watching GPU/CPU frame times
5) Once you find that place when CPU frame time is spiking, take note where it is.

On my 4770K with HT, when looking into places with reduced geometry load (but not too low) - like a coast + sky + sea. I was getting hit with 70ms CPU times. That is very quantifiable and the same was happening in MP, creating detectable jitter.


Now disable core parking (forgot if reboot is required, but do so). And comeback to same damn place in BF4. Poof, CPU spikes are gone and FPS is silky smooth (and ffs it should be, there is not AI, MP code load or any substantial movement in that frame).

So to sum it, core parking is EVIL in Win7 and creates a very detectable frame jitter, 70ms frame time is eternity in BF4 (and what is funny, resulting average FPS is still great, in a region of 55 FPS in case of vsync.

 

[ozzy702]

That was my experience. Also, regarding HT not helping with BF4, I find that suspicious to say the least. Too much info already out there, including testing done here by us forums users, that HT and more threads benefits BF3 and BF4 significantly.
I suspect, due to all the conflicting info regarding core parking, windows versions etc, that something wonky is going on in an inconsistent way with BF4. Its just too early to know for sure whats going on IMO.

Agreed. When I went from a 2500k to a 2600k I noticed a massive improvement in overall frame rates, especially so when looking at minimums.

BF4 may be different, I have no way to tell since I am on a 3770k box now and have no other Intel box to play with.

 

[toyota]

Agreed. When I went from a 2500k to a 2600k I noticed a massive improvement in overall frame rates, especially so when looking at minimums.

BF4 may be different, I have no way to tell since I am on a 3770k box now and have no other Intel box to play with.

sorry but I have a hard time believing that and I think most of that was in your head. Crysis 3 is probably the only game that shows any playable difference with HT. Most of the games will not see any difference at all and certainly not any playable difference.