Only 2 cores seem to be doing the work?

[chuggs1977]

Hi all i play world of warcraft and was thinking that i should be getting better fps and such, usually i get around 60fps more less solid barring been in orgrammar which it then goes down to around 25-30fps all settings are on high,also when i have looked at my taskmanager out of the 6 cores i have 2 seem to be doing a lot of work and the other 4 hardly anything, im new to all this technical side of things really so dont know if this is right or if something is holding my system back ? This is the spec of my machine, and also do i need to update the processer drivers looking at them there sayng the driver is from 2006 but never had to update processer drivers before ?

Asus sabertooth 990fx motherboard
Phenom 1100T black edition x6 @ 3.30ghz/oc 3.80ghz
2 x GTX 560ti (448)
2 x 64gb ssd
1x tb harddrive
1000w cooler master silent
Fractal design R3 case in Artic White
64 bit windows ultimate
16gb Corsair xms ram 1600mhz

Also i never run this game in sli because i never see any difference in fps but im pretty sure it should be running smoother than what it is any help in this matter would be greatly appreciated.

 

[unoriginal1]

the majority of games only utilize 2 cores. That's why everyone says the i5's are the best procs for gaming.

Games do not use Hyper threading or more then 4 cores. (usually only 2 depending on the game)

 

[luciferano]

Are you using a DX11 version of World of Warcraft? If not, then doing so might help.

Either way, most games, especially old games such as WoW, are not able to use many cores/threads effectively. One or two well-utilized threads and a few other much lighter threads is very common. This is a huge part of why AMD is often seen as underperforming compared to Intel even though AMD tends to have much higher performance at any given budget through sheer core count.

 

[luciferano]

the majority of games only utilize 2 cores. That's why everyone says the i5's are the best procs for gaming.

Games do not use Hyper threading or more then 4 cores. (usually only 2 depending on the game)

That's not entirely true. A lot of games use more than two cores. Most games that do use many cores simply don't use more than two very efficiently (IE going from one to two cores might double performance if all else is equal, but going from two to three or four is unlikely to add more than around 20-30% more performance in most games). Also, there are some games (very few) such as BF3 MP that can actually use six or even eight cores almost perfectly. Almost all modern games can use Hyper-Threading. This is supported by the OS, not the game. The game sees then as regular cores and going from a Pentium (two cores) to an i3 of the same micro-architecture and frequency (two cores with Hyper-Threading) is known to be a considerable performance improvement (20-30%) in many modern games.

 

[unoriginal1]

That's not entirely true. A lot of games use more than two cores. Most games that do use many cores simply don't use more than two very efficiently (IE going from one to two cores might double performance if all else is equal, but going from two to three or four is unlikely to add more than around 20-30% more performance in most games). Also, there are some games (very few) such as BF3 MP that can actually use six or even eight cores almost perfectly. Almost all modern games can use Hyper-Threading. This is supported by the OS, not the game. The game sees then as regular cores and going from a Pentium (two cores) to an i3 of the same micro-architecture and frequency (two cores with Hyper-Threading) is known to be a considerable performance improvement (20-30%) in many modern games.

Hence why I said most.

*edit* Also Hyper Threading does more bad then good. You wouldn't see a 20-30% increase in performance. Hyper threading can cause stutters in game play. No game besides bf3 to my knowledge even uses it. Possible down the road it would help but as for todays games. it decreases performance when compared to an i5.

 

[unoriginal1]

But to be fair... I guess if we did want to get technical. To say it *doesn't use* hyper threading. That would be inaccurate. So my bad on poor choice of wording.

 

[gamerk316]

To be fair, the lack of a single application to scale on multiple CPU's was ONLY discovered way back in the 80's. I really am surprised that people expect a single application to scale well, when the very design of the hardware inhibits this.

Seriously, most programs are serial in nature. So right there you lose the ability to scale well. Then you have to be worried about synchronization, hardware I/O, and the like. And so on and so forth.

 

[geekapproved]

Actually there is a game or two benefit from hyperthreading.

Don't ask me which ones though.

Some people say you need 4 cores to play BF3 with good fps, I say horsecrap, I own noobs all day long and I run a i3 and OC HD5850. Windows does a good job of spreading load across all 4 threads.

 

[Wisecracker]

In *some* gaming, booting with 2 cores (instead of six) will see a nice bump as they better share the L3 cache.

You will certainly see a nice bump by cranking your IMC/NB to the 24-2500MHz range. For each 10% you increase the speed of the IMC/NB above stock 2000MHz, memory bandwidth is increased 3-4% and latency is reduced 3-4%.

 

[InvalidError]

That's not entirely true. A lot of games use more than two cores. Most games that do use many cores simply don't use more than two very efficiently.

Do not confuse Windows scheduling the game's few active threads round-robin or randomly across 4+ threads/cores as "using many cores". You could run single-threaded application on a 6C12T CPU and see activity on all 12 "cores" simply due to scheduling but the overall CPU utilization in Task Manager would only be 8.3%, equivalent to a single logical core at 100% load.

If you have a game that maxes out at 40% CPU load in Task Manager on a quad-core CPU, it is effectively using only 1.6 cores.

 

[luciferano]

Do not confuse Windows scheduling the game's few active threads round-robin or randomly across 4+ threads/cores as "using many cores". You could run single-threaded application on a 6C12T CPU and see activity on all 12 "cores" simply due to scheduling but the overall CPU utilization in Task Manager would only be 8.3%, equivalent to a single logical core at 100% load.

If you have a game that maxes out at 40% CPU load in Task Manager on a quad-core CPU, it is effectively using only 1.6 cores.

I did not make any such mistake. What I said was correct. A lot of games have a few threads that do some work, just not a lot of work and can't efficiently utilize cores as much as one or two *main* threads. This is a common problem caused by developers not coding effectively for high amounts of threads.

 

[unoriginal1]

I apologize to the OP as I'm partially to blame for the tangent .

But does anyone know how to improve his game play on WoW?

 

[luciferano]

Like I said at the start of the thread, the best thing to try would be making sure that OP is using the DX11 launcher for WoW. I don't think that there is much else that can be done about the situation. You can't force a game to use more threads than it has unless there is support for it that is locked that can be unlocked.

 

[mikes1992]

I highly doubt this is possible but I've always wondered if it's possible to combine cores into a single virtual thread and say running 2 3ghz cores at a logical speed of 6ghz... they'd need to share resources though so they would be using the same L1 and L2 cache (I assume?)... I believe it would be near impossible to do though.

Also is it possible to run an octave frequency over a CPU? like for example if you got an extra 1hz every 10hz so a 3ghz core would get 300mhz extra speed? would this let you achieve slightly higher frequency's with a little less heat generation?

I'm just speculating/sharing thoughts haha.

 

[InvalidError]

I did not make any such mistake. What I said was correct.

While you may be correct, there are tons of people erroneously concluding that an application is using more cores than it actually does only because they see activity across multiple cores when that activity is only an artifact from the scheduler playing musical chair with the one or two active threads.

Personally, an application that has 20 threads but only two that do any meaningful amount of work is fundamentally targeted for dual-core CPUs.

 

[luciferano]

I highly doubt this is possible but I've always wondered if it's possible to combine cores into a single virtual thread and say running 2 3ghz cores at a logical speed of 6ghz... they'd need to share resources though so they would be using the same L1 and L2 cache (I assume?)... I believe it would be near impossible to do though.

Also is it possible to run an octave frequency over a CPU? like for example if you got an extra 1hz every 10hz so a 3ghz core would get 300mhz extra speed? would this let you achieve slightly higher frequency's with a little less heat generation?

I'm just speculating/sharing thoughts haha.

Combining cores like that doesn't really work. I think that I could figure out something if you want to get into it, but with current CPUs, that simply doesn't work.

Maybe frequency ticks like that could be done without major architectural changes.

 

[luciferano]

While you may be correct, there are tons of people erroneously concluding that an application is using more cores than it actually does only because they see activity across multiple cores when that activity is only an artifact from the scheduler playing musical chair with the one or two active threads.

Personally, an application that has 20 threads but only two that do any meaningful amount of work is fundamentally targeted for dual-core CPUs.

I see what you're getting at. For measuring core count efficiency in games, I prefer changing the game's core affinity so that it can only run on a specific core, then two, then three, and so on so that I can get what should be a more accurate measure of how the game scales with more cores (or threads for Hyper-Threading capable CPUs). Do you think that this is an inaccurate method?

 

[nforce4max]

When it comes to wow you just not going to get the best fps but that 6 core cpu can handle a lot of wow clients at once. I get similar fps as you but when I open up more clients and begin to log several accounts fps remains the same. Wow uses an ancient engine so what you got is about right. Keep in mind that wow doesn't scale much at all in sli/crossfire.

 

[InvalidError]

Do you think that this is an inaccurate method?

Playing with affinity does have the benefit of highlighting exactly how many bursty simultaneous things the game might be throwing around on different threads.

For applications/games that use "bursty threading", there might be a noticeable difference between this and my task-manager evaluation. Since few programs are written this way due to the complexity and overhead of setting things up that way and reconciling/synchronizing results afterward, Task Manager CPU% is still a simple and reasonable approximation of how many cores are used most of the time.

 

[luciferano]

Playing with affinity does have the benefit of highlighting exactly how many bursty simultaneous things the game might be throwing around on different threads.

For applications/games that use "bursty threading", there might be a noticeable difference between this and my task-manager evaluation. Since few programs are written this way due to the complexity and overhead of setting things up that way and reconciling/synchronizing results afterward, Task Manager CPU% is still a simple and reasonable approximation of how many cores are used most of the time.

Let's assume that you have a quad-core CPU. 40% might mean 1.6 cores utilized, but it might be that one core is utilized and another is about half utilized, or it might mean that one core is about 80% used and two others are about 40%. It seems less accurate because it doesn't really tell you how the game uses the CPU.

 

[InvalidError]

Let's assume that you have a quad-core CPU. 40% might mean 1.6 cores utilized, but it might be that one core is utilized and another is about half utilized, or it might mean that one core is about 80% used and two others are about 40%. It seems less accurate because it doesn't really tell you how the game uses the CPU.

But there is no guarantee that you will see any difference between dual or quad core since the dual core should still have more than enough processing power to handle the effective workload of 1.6 cores.

I personally would not lose sleep over this. The bottom line is I would not expect much of a speed-up going from dual to quad with software that only has an effective load of 1.8 cores on quad but I know for sure to expect a slow-down downgrading to a dual-core if the effective load on the quad is higher than two cores. That was what my "TM test" is about.

 

[luciferano]

You have a point, but the problem is that it might not translate to similar usage with different CPUs. For example, let's assume that we've been talking about a Core 2 Quad. Applying the same workload to say an FX-4100 or an i3 might have different results. For example, the FX-4100's first core of each module is faster than the second core if the second core has lower utilization than the first core.

The same is true if the second core has higher utilization, but the point is the same. The i3 uses Hyper-Threading that with highly threaded performance, let's it more or less keep up with many lower end quad cores such as most Core 2 Quads and some Phenom II x4s and FX-4100 with the Ivy i3s, so a load of say 1.8 cores on the Core 2 Quad might be more like 1.3 on the i3 or depending on the workload, more like 2.1 or whatever. My method gives you real scaling comparisons that should hold true even with CPUs of different micro-architectures and performance characteristics whereas your depends on the individual CPU.

 

[InvalidError]

You have a point, but the problem is that it might not translate to similar usage with different CPUs. Applying the same workload to say an FX-4100 or an i3 might have different results.

Hello Mr Obvious.

The performance of AMD chips cannot be compared with Intel's in any meaningful way in this context. The usage of similar architecture and clock rates was implied all along.

 

[luciferano]

Hello Mr Obvious.

The performance of AMD chips cannot be compared with Intel's in any meaningful way in this context. The usage of similar architecture and clock rates was implied all along.

Even comparing different Intel CPUs could make your method fail. Heck, comparing two CPUs that are identical in all but frequency can make it fail. Mine has no such fallacy.

 

[chuggs1977]

Hi all and thanks for the replies,some of the replies just blow my mind lol as im new to all this.Couple of things am i best enabling hyperthreading or leaving it disabled which is what its on at the moment,also i run the game using directx11 which i enabled in game is this right too ? or do i have to enable it before i boot the game up somewhere? also i read somewhere that when computer boots up to change the boot up core to the max i have which is 6 is this ok ? thanks again and sorry for the noobish questions but im getting there