Force GTA IV to utilize multiple Cores

[stormmaster]

Hello,

is there a way to force GTA IV to use multiple cores?

because i think that although i have 8 CPU cores (i7 920 with enabled HT) GTA IV only use one of them (only 17% CPU usage in the benchmark and only 23FPS no matter which graphic settings i use)

 

my pc specs are:

Intel i7 920 @2.6ghz

12GB RAM

Radeon HD 5870 2GB (Catalyst 10.11)

Win 7 Prof 64bit

GTA IV 1.0.7.0

 

tia

[mkey82]

No, you can't. GTAIV is three core/threads "optimized" and the only way around it is to up the CPU clocks, i.e. overclock it.

[stormmaster]

my problem is that it seems that gta only run on one core of eight. if it's run on three core/threads there should be a cpu load of at least 37.5% not only 17 % (i think 12.5% of the game rest form background processes)

[mkey82]

Benchmark is a broken tool and it shouldn't be taken seriously. If you want more dependable results, you'll have to use some other program, like everest or at least taskmanager to estimate the actual processor usage.

[IronHide-AW]

 

my problem is that it seems that gta only run on one core of eight. if it's run on three core/threads there should be a cpu load of at least 37.5% not only 17 % (i think 12.5% of the game rest form background processes)

The best way I've found for I7's so far to see true CPU ramping (and therefore usage) in real time is TMonitor64 (or TMonitor if you on 32 OS).

 

You can find it right here.

 

It's an excellent simple tool showing the frequency of each core in a lapse graph you can log to disk. You can just use the visual graph if you're testing something that isn't full screen but for game just use disk logging option.

 

So, run that, turn on logging, run the game and do stuff (the more intense the better. Rampage the city for a bit) then exit the game and turn off logging.

 

When logging is turned off the log file will then be created and you can see core ramping behavior. A core ramping to its maximum allowed boost = a core at 100% usage so you can just do simple math on your own numbers and know your real usage over time with the game.

 

Some examples I just did real quick for this post. Entered game and drove for about 20 seconds then exited. Note I'm 6 cores so you'll see 4 cores instead but here are some random entries from the log I snatched out:

 

Each number set is the frequency a core reached via boost because it was being taxed by the game.

 

2272.4;1604.0;1604.0;4010.1;3208.1;1737.7; = 4 cores were taxed, 2 minor, 1 medium, 1 major.

2272.4;1604.0;1604.0;4010.1;3341.7;1604.0; = 3 cores were taxed, 1 minor, 1 medium, 1 major.

2539.7;2138.7;1604.0;3876.4;3609.0;1604.0; = 4, 2 min, 2 med.

1604.1;4143.9;1604.1;1604.1;1604.1;4143.9; = 2, 2 maj.

2539.6;2005.0;1604.0;4009.9;3341.6;1604.0; = 4, 2 min, 1 med, 1 maj.

2272.4;1604.0;1604.0;4143.8;3208.1;1604.0; = 3, 1 min, 1 med, 1 maj.

etc etc.

 

Obviously these will include any load put on the cores by OS or apps in the background that spike or otherwise do something to cause an increase but even with that fact it is extremely useful IMO to see what you're asking. It's also the perfect way to confirm TDP overclocking is working. Speaking of that this reminded me I need to increase again I brought it down to 4.1 the other day while doing performance analysis of Metro 2033 and wound up playing and forgetting I did that.

 

Edited:

Fixed descriptions. Was distracted with conference call coming and did it in haste. = remarks are now more accurate for their entries.

Edited December 7, 2010 by IronHide-AW

[Jigglyass]

Just so you know, i7 does not have 8 cores, they have 4 cores and 4 logical cores = 4cores/8threads.

Thread =/= a core.

HT is useless, it actually hurts performance in some games like L4D1 or 2.

Id turn it off, CPU would generate less heat and OC better on less volts.

Gta is programed to utilize only 3 cores, and it will not utilize threads.

[Ron Taboga]

I noticed when I switched from my old Athlon II X4 to the Phenom II X6 that IV runs noticeable faster. It may be only the L3 Cache, but i don´t think so.

 

The Game can only use 3 Cores in an optimal way, but more cores are utilized minimally i guess. GTAIV also runs better on 4 Cores than on 3, also on 6 than on 4. But this seems to be only a minor performance win (max. 5-10 FPS).

 

The biggest Performance Bang ist still between an 2 Core and 3 Core CPU thats right.

[mkey82]

 

The Game can only use 3 Cores in an optimal way, but more cores are utilized minimally i guess.

In a few words, no.

[Ron Taboga]

The Game can only use 3 Cores in an optimal way, but more cores are utilized minimally i guess.

In a few words, no.

Than it was maybe the Advantege of the L3 Cache in my case, and maybe the 250 MHz more than on the Athlon II X4.

 

So GTA IV runs the same with a 3 Core and 4 Core CPU when they have the same clock and architecture?

[Jigglyass]

The Game can only use 3 Cores in an optimal way, but more cores are utilized minimally i guess.

In a few words, no.

Than it was maybe the Advantege of the L3 Cache in my case, and maybe the 250 MHz more than on the Athlon II X4.

 

So GTA IV runs the same with a 3 Core and 4 Core CPU when they have the same clock and architecture?

In most scenario, yes.

[mkey82]

Yes, I'd say that should be true. The only thing that can make the difference is the fact GTAIV is not running alone on the system, so effectively the fourth core may be used by other running programs. Without the fourth core, the system will have to take away some CPU time from GTAIV thus lowering the performance. Not considerably, though.

 

Tests have been done, just disable one core or set core affinity for GTAIV to 3 cores and check it out. Only when you disable more then one core you will see drops in performance.

[stormmaster]

i seems that gta has problems with the hyperthreading.

after i disabeld ht the gta benchmark runs with an avarage of 69 fps and has a cpu load of 71%

[IronHide-AW]

^ there are a number of games that wind up that way since they don't take advantage of HT having it enabled can in fact hurt performance because it probably means gates and other stuff are actively tested but not used so it just induces micro delays instead of actual performance boost if was being used. I certainly remember from so long ago when HT came-to-be it was from Intel engineers realizing the original architecture was done in a way that meant the maximum possible use of the CPU was actually at 50% not 100%. So they realized they could add a way to execute on the up or down side of a wave too and use the remaining 50% in a cycle to get a real 100% CPU utilization hence the birth of HT. It's been very long so that's probably pretty rusty but that's what I recall anyway. Haven't looked up HT in a long damn time though.

[Jigglyass]

^ there are a number of games that wind up that way since they don't take advantage of HT having it enabled can in fact hurt performance because it probably means gates and other stuff are actively tested but not used so it just induces micro delays instead of actual performance boost if was being used. I certainly remember from so long ago when HT came-to-be it was from Intel engineers realizing the original architecture was done in a way that meant the maximum possible use of the CPU was actually at 50% not 100%. So they realized they could add a way to execute on the up or down side of a wave too and use the remaining 50% in a cycle to get a real 100% CPU utilization hence the birth of HT. It's been very long so that's probably pretty rusty but that's what I recall anyway. Haven't looked up HT in a long damn time though.

TL;DR;

HT is useless in games and actually hurts performance.

[IronHide-AW]

Then why quote it, increasing the thread length even more? Also, saying HT is useless in games applying it generally across everything is both funny and untrue.

 

Edited:

 

To hopefully prevent further useless responses founded on absolute nonsense I advise first chew this a bit.

 

Also, I've personally done experiments in hobby as well as number of professional projects that use SMP and it's awesome when used correctly. Hell you could even go get my Mass Effect 2 Coalesced Editor and if I recall I added a threaded file DIFF that explicitly shows the tremendous benefit of SMP, again, where its applicable.

 

Edited December 9, 2010 by IronHide-AW

[mkey82]

I believe when he said "useless" he meant practically useless as most games today are not that hard on the CPU anyway as they just don't need that much CPU power or are just lame console ports that you could write a book about and not use words "used correctly" one single time.

 

However, stormmaster here probably has some unrelated issue as he's the first person I hear from having this problem. There is really no logical reason HT would actually cause any damage.

[JohnZS]

GTA IV loves multiple cores, for those of you who doubt it please see Core i7 980X review

Notice the MINIMUM FPS on the Intel HEX core.

John

[IronHide-AW]

I believe when he said "useless" he meant ...

... what he wrote, which is a dismissive insult on my taking the time to describe something and then posting a blanket statement with arrogance behind it. But to be arrogant it usually helps to be right, because when you're not you just wind up looking like a moron instead. I'd say that describes that blanket-wrong statement he made pretty well. But I digress and say multi-core and HT code is not only here to stay but a revolution in the making, especially for AI as an example. There's good stuff ahead

[mkey82]

There's good stuff ahead

I think you're in for a long wait.

[IronHide-AW]

I think you're in for a long wait.

And you should be more optimistic.

[mkey82]

It has absolutely nothing to do with my disposition.