Crossfire 7950 only 60% gpu usage

[grandpatzer]

So I'm trying out my new 120hz display 1080p, obivously 2 AMD 7950 should be able to run 120fps correct? NO!

I started Alan Wake American Nightmare and to my sad suprise the FPS is hovering 50-95fps, mostly is is about 75fps-85fps.

I the odd thing is that the CPU hovers around 50% while GPU's hover around 60%, I see better smoothness as I'm upgrading from 60hz display but I really wanted to get the 120fps+ that I can get with GPU load at 90% +

 

[runs2far]

SLI/Crossfire does not always work out of the box, spend some time figuring out why it does not work.

 

[magoo]

Do you have the latest CAPs loaded?

You will find that in some games, Crossfire more than Surround is woefully poor.
Trust me I've had the same gripe about Crossfire.

Disable Crossfire and see what you get?

 

[raghu78]

have the latest drivers and CAP installed. 12.8 WHQL and 12.7 CAP 3. check performance with CF turned off and on as suggested above.

 

[grandpatzer]

have the latest drivers and CAP installed. 12.8 WHQL and 12.7 CAP 3. check performance with CF turned off and on as suggested above.

This is what I have, also I used to have eyefinity setup with 85-100% gpu load, so now having 60% on both gpu's is sad.

So I should turn off CF then start the game and then turn it on again?

For what it's worth I know CF is active this I'm 100% certain of.

 

[magoo]

This is what I have, also I used to have eyefinity setup with 85-100% gpu load, so now having 60% on both gpu's is sad.

So I should turn off CF then start the game and then turn it on again?

For what it's worth I know CF is active this I'm 100% certain of.

I know how you feel. I've seen this more often than not.
If you look at the half-full glass though, the GPUs are not working hard to provide your graphics.

Is it just the frame rate that bothers you or are you seeing poor performance?

Do you see similar utilization in other titles?

Starting and stopping CF won't help you. There are just some of the profiles that are bad.
If you are into experimentation, you could always try a different user defined CF profile from the make your own portion of the CCC.

 

[Final8ty]

What is your CPU running at as it looks like CPU bottleneck to me.

 

[grandpatzer]

I know how you feel. I've seen this more often than not.
If you look at the half-full glass though, the GPUs are not working hard to provide your graphics.

Is it just the frame rate that bothers you or are you seeing poor performance?

Do you see similar utilization in other titles?

Starting and stopping CF won't help you. There are just some of the profiles that are bad.
If you are into experimentation, you could always try a different user defined CF profile from the make your own portion of the CCC.

Well honestly the game play is alot better compared to 60hz display, more smooth with my new 120hz display.

Still it's not buttersmooth, In counterstrike: GO I get smoother gameplay it is 190fps and each gpu is only 35% gpu load so not wery demanding game but looks good

What is your CPU running at as it looks like CPU bottleneck to me.

I have a Intel i5 2310 overclocked to 3.4ghz quad, it is similar performance to a i5 2500k stock.

Also I have hwinfo64 and msi AB showing real time CPU load and it is only 50% cpu load

 

[psyside]

So you cheap out on like 40 dollars and yet you bought 700$ cards?

Great.

Anyway, go back to 12.7 and latest CAP.

 

[grandpatzer]

So you cheap out on like 40 dollars and yet you bought 700$ cards?

Great.

Anyway, go back to 12.7 and latest CAP.

Not that it is any of youre business but I bought the CPU used for 50% market price, either way what benefit do I get from buying 1000$ intel cpu when only thing it does is decrease my cpu load from 50% to 30%.

Maybe 12.7 will help, doubt it..

 

[sMiLeYz]

Sounds like you're CPU limited. I have the similar thing happen to me, I was running 460's in SLI with an Intel Q9550 quadcore. I was dead set on not upgrading until intel released hexa-cores or octo-cores than I got a sweet deal on 120hz monitor.

I was sad that I was pushing 70-80 fps at max, so I just went all out and bought an Sandybridge. Overclock that sucker and I was hitting 100's finally. I just picked up a 7970 and now im starting to see 120fps more steadily.

 

[magoo]

Not that it is any of youre business but I bought the CPU used for 50% market price, either way what benefit do I get from buying 1000$ intel cpu when only thing it does is decrease my cpu load from 50% to 30%.

Maybe 12.7 will help, doubt it..

I have found Cat 12.4 to be the only reliable driver of the 12 series so far.
Give it a shot combined with the 12.7 CAPs.

 

[psyside]

Not that it is any of youre business but I bought the CPU used for 50% market price, either way what benefit do I get from buying 1000$ intel cpu when only thing it does is decrease my cpu load from 50% to 30%.

Maybe 12.7 will help, doubt it..

I didnt say you need to buy 1000$ cpu, i said you need to buy 220$ CPU huge difference there, its called balanced configuration, overclocked CPU with high end multy gpu setup is a must.

But if you get it for so cheap its ok, try 12.7 with latest cap.

P.S. Chek out if your turn Vsync on, reset catalyst to defaults, then try again.

 

[grandpatzer]

I just remembered that I never upgraded to 12.8 because I read loads of people having BSOD.

I actually now went from 12.7 + CAP 12.7 3 to ----> 12.8 + CAP 12.7 3 and now there is noticable FPS increase, still I see 75fps and a bit lower when at lowest.

I'm still not convinced CPU upgrade will help me as the CPU is only 50% load, incase the CPU is not able to have enought power should it not be 100% cpu load?

 

[Sodapopjones]

I just remembered that I never upgraded to 12.8 because I read loads of people having BSOD.

I actually now went from 12.7 + CAP 12.7 3 to ----> 12.8 + CAP 12.7 3 and now there is noticable FPS increase, still I see 75fps and a bit lower when at lowest.

I'm still not convinced CPU upgrade will help me as the CPU is only 50% load, incase the CPU is not able to have enought power should it not be 100% cpu load?

Loading your CPU and the actual speed of your CPU are two different things lol. Having multiple threads loads your core, not how fast the clock speed is.

If I run prime95, my CPU is going to see 100% load, but that doesn't mean I'm going to run crossfired 7950's lol.

 

[grandpatzer]

Loading your CPU and the actual speed of your CPU are two different things lol. Having multiple threads loads your core, not how fast the clock speed is.

If I run prime95, my CPU is going to see 100% load, but that doesn't mean I'm going to run crossfired 7950's lol.

well running Alan Wake AN I have locked my CPU to 3.4ghz on all 4 cores, the CPU core load on each core ranges from 30-85% on each core and 50% CPU usage on average.

yes Prime95 can get 100%, Alan Wake AN can not get 100% cpu load, I fail to see how a more expensive CPU will get me more fps?

 

[magoo]

well running Alan Wake AN I have locked my CPU to 3.4ghz on all 4 cores, the CPU core load on each core ranges from 30-85% on each core and 50% CPU usage on average.

yes Prime95 can get 100%, Alan Wake AN can not get 100% cpu load, I fail to see how a more expensive CPU will get me more fps?

There's a pretty good article on just that circulating around the H archives.
I want to say it was in relationship to the GTX 580s and compared the test bench using the standard quad core at 3.6 GHz on an X58 platform to a newer Sandy Bridge 3770k on a Z77 workstation platform.

The only thing that changed was the CPU and the performance of the GPUs jumped dramatically.

 

[grandpatzer]

There's a pretty good article on just that circulating around the H archives.
I want to say it was in relationship to the GTX 580s and compared the test bench using the standard quad core at 3.6 GHz on an X58 platform to a newer Sandy Bridge 3770k on a Z77 workstation platform.

The only thing that changed was the CPU and the performance of the GPUs jumped dramatically.

Well I found it, also it is 5760*1080 that requires more CPU power compared to 1080p.
While there is a difference I would not say it is enormous, also the reviewer does not explain what the CPU load % was,

Also while my high overclocked 2x 7950 at 1100/1575 should perform about 10% more compared to a stock 7970, I would assume 3 stock GTX 580 may be more powerful compared to my 2x 7950.

 

[rUmX]

Just because you see 50% CPU usage doesn't mean you don't have a CPU bottleneck. All it means is basically that your system is using at most 2 cores (of a 4 core cpu). Most apps and games today don't use more than 2 threads at most. In this case clock speed is king. I gained around 15-25fps just going from stock 2500K to overclocked at 4.5ghz in Starcraft 2.

 

[grandpatzer]

Just because you see 50% CPU usage doesn't mean you don't have a CPU bottleneck. All it means is basically that your system is using at most 2 cores (of a 4 core cpu). Most apps and games today don't use more than 2 threads at most. In this case clock speed is king. I gained around 15-25fps just going from stock 2500K to overclocked at 4.5ghz in Starcraft 2.

but that does not make sense, in windows 7 my 4 cores are around 1-5% load, when I start up Alan Wake AN then each core usage is between 30-80%.

If indeed Alan Wake AN was a dual core game and bottlenecked by a i5 3.4ghz quadcore then 2 cores should be 100% while the 2 other cores would be 1-5% load

 

[runs2far]

OK, this is getting semi interesting.

Applications with less threads than available cores can get bounced around, making it look like there is plenty of CPU power when the application is actually CPU limited.
Try locking your game to one, two and then three cores and see if there is any performance difference.

http://www.addictivetips.com/window...cessor-affinity-to-an-application-in-windows/

 

[sirmonkey1985]

What is your CPU running at as it looks like CPU bottleneck to me.

yeah i agree with this one.. you're either cpu limited or the game engines a piece of crap and doesn't support more then 2 cores which then would mean you are going to be cpu limited with a 1045T.

but that does not make sense, in windows 7 my 4 cores are around 1-5% load, when I start up Alan Wake AN then each core usage is between 30-80%.

If indeed Alan Wake AN was a dual core game and bottlenecked by a i5 3.4ghz quadcore then 2 cores should be 100% while the 2 other cores would be 1-5% load

its windows load balancing, even though you see it using all the cores available its still only uses the same resources across all 4 cores that it would use if the system was a dual core. limit the game to 2 threads and you'll see the exact same performance except with 2 cores at full load and honestly this would help you since it should trigger the boost option in the 1045T if you are running stock.

 

[runs2far]

yeah i agree with this one.. you're either cpu limited or the game engines a piece of crap and doesn't support more then 2 cores which then would mean you are going to be cpu limited with a 1045T.

The OP has a i5 2310 and an outdated sig.

 

[grandpatzer]

OK, this is getting semi interesting.

Applications with less threads than available cores can get bounced around, making it look like there is plenty of CPU power when the application is actually CPU limited.
Try locking your game to one, two and then three cores and see if there is any performance difference.

http://www.addictivetips.com/window...cessor-affinity-to-an-application-in-windows/

So I start Alan Wake AN then press windows button and put affinity to 2 cores?
But if the game is using 4 cores as Hwinfo64 suggests this won't prove anything(?)
Or if the game uses 4 threads then fps should be lower with 2 core (?)

According FAQ it is indeed quad core game linky

For a solid performance Remedy recommends a modern quad-core CPU and a graphics card with around 2000 in 3DMark Graphics Score as seen in here: http://community.futuremark.com/hardware/gpu

So unless Remedy lies, indeed my cores are using 30-80% each and barely breaking 50% cpu overall load.

 

[runs2far]

So I start Alan Wake AN then press windows button and put affinity to 2 cores?
But if the game is using 4 cores as Hwinfo64 suggests this won't prove anything(?)
Or if the game uses 4 threads then fps should be lower with 2 core (?)

According FAQ it is indeed quad core game linky


So unless Remedy lies, indeed my cores are using 30-80% each and barely breaking 50% cpu overall load.

Remedy recommends a quad core for solid performance, they do not write "this game utilizes four cores". Even if it does use four cores, you do not know how well it scales.
The game may have a primary thread that is CPU limited and bounces between the cores and an unknown number of threads that sit halfway idle waiting for the primary thread to dish out some work.

If you start the game and force it to stay on two cores, you may actually know of it is able to max out those two cores, which is a hint at being CPU limited.
If you then notice an increase in frames by going to three and later four cores, you would know that it actually scales to four cores.

This is not a solution, it is an aid for us in determining what is limiting your game performance.

 

[Final8ty]

Keep this simple, let's say that for whatever reason the game code is only using one in four cycles so your CPU is only at load 25% for the gfx update let's say at 1Ghz, if you run your CPU at 2Ghz and it's still one in four cycle then you would of still increased the info to the gfx card giving you more GPU load and fps.

 

[grandpatzer]

Putting only 2 cores to Alan Wake: AN I get 80% cpu load on the 2 active core each
with 2 cores I get 109fps scene1 and 80fps scene2.
with 4 cores I get 120fps scene1 and 90fps scene2.

So I guess the game just uses up to 80% of each core, so I could aswell run this game on a 2.7Ghz i5 quadcore then.

My conclusion is that the 50% cpu load is true and that the game is not made for powerful systems with crossfire 7950 and a sandy bridge cpu.

 

[Final8ty]

Putting only 2 cores to Alan Wake: AN I get 80% cpu load on the 2 active core each
with 2 cores I get 109fps scene1 and 80fps scene2.
with 4 cores I get 120fps scene1 and 90fps scene2.

So I guess the game just uses up to 80% of each core, so I could aswell run this game on a 2.7Ghz i5 quadcore then.

My conclusion is that the 50% cpu load is true and that the game is not made for powerful systems with crossfire 7950 and a sandy bridge cpu.

But again my example still applies because it still 1 in 4 cycles per core which would be faster on the quad than on the dual because its still more data handled per second per cycle and over clocking will still gain you more.
You don't have to be at 100 loads on the CPU before you gain from over clocking it if the bottleneck is at the CPU.
Some things can use every cycle because they are simple, others cant so to make up for that you run your cycles faster.

I remember running my men at 1T and switched too 2T which is every other cycle, i lost 10fps, i then OCed the mem and got back the 10fps.

 

[grandpatzer]

But again my example still applies because it still 1 in 4 cycles per core which would be faster on the quad than on the dual because its still more data handled per second per cycle and over clocking will still gain you more.
You don't have to be at 100 loads on the CPU before you gain from over clocking it if the bottleneck is at the CPU.
Some things can use every cycle because they are simple, others cant so to make up for that you run your cycles faster.

I remember running my men at 1T and switched too 2T which is every other cycle, i lost 10fps, i then OCed the mem and got back the 10fps.

I totally forgot about memory, I think that in my case memory tweaking will not increase min fps(?)

I actuall might have been wrong about cpu bottleneck, as when I look at MAX cpu load all cores are 99%, but when I open the logged file the cores are 30-80%, so I guess the game for most uses just 2 cores.

Here is a snap shot what hwinfo log data looks like:
gpu1 .. 2.. core1 ... 2..... 3 ..... 4 ....
63,9 63,9 65,7 21,4 45 65,7

Also it's same cores that are 60-70% while the 2 other are around 30-40%.

 

[Final8ty]

There was another user Flanno on another forum who had the exact same cards as me and in one game he had double the FPS as he was using an OCed Intel i7-990X and i was using a quad AMD at 3.6. and my CPU was only at 50% load.

Phenom II X6 1090T 4Ghz NB 3Ghz that im using now still bottlenecks me at times and its never at 100% load unless it loading levels in BF2/BF3.

 

[runs2far]

Memory will not change your performance when you have an i5.

If the game does scale to four cores you have wasted cores and it does not matter if you have 4 or 16 cores if the game does not use more than 2.

Looking at your numbers it looks like the game barely uses 3 cores.
With the 80% load on two cores I'm really leaning towards it's the GPU drivers or CF profile that are failing

I should have suggested this earlier, yes I am tired of this CPU limited Bsh*t too.
You clearly know what you are doing with that CPU, try clocking it down to 3 GHz and keep turbo core disabled.
If you get about 10% less FPS when the CPU is running at 3 GHz, it's CPU limited.

 

[grandpatzer]

Memory will not change your performance when you have an i5.

If the game does scale to four cores you have wasted cores and it does not matter if you have 4 or 16 cores if the game does not use more than 2.

Looking at your numbers it looks like the game barely uses 3 cores.
With the 80% load on two cores I'm really leaning towards it's the GPU drivers or CF profile that are failing

I should have suggested this earlier, yes I am tired of this CPU limited Bsh*t too.
You clearly know what you are doing with that CPU, try clocking it down to 3 GHz and keep turbo core disabled.
If you get about 10% less FPS when the CPU is running at 3 GHz, it's CPU limited.

I put in bios the CPU to 2.7Ghz, so it is 20% less power or 80% of 3.4Ghz.

A segment of the game on average the 2.7Ghz made the Gpu's have 55% gpu load.
Same segment of the game on average 3.4ghz made the Gpu's have 60% gpu load.

When lowest fps 3.4Ghz gave: 67fps
When lowest fps 2.7Ghz gave: 55fps

So I guess the game is not optimized for quadcore then and a stronger cpu gives more fps.

 

[Final8ty]

I put in bios the CPU to 2.7Ghz, so it is 20% less power or 80% of 3.4Ghz.

A segment of the game on average the 2.7Ghz made the Gpu's have 55% gpu load.
Same segment of the game on average 3.4ghz made the Gpu's have 60% gpu load.

When lowest fps 3.4Ghz gave: 67fps
When lowest fps 2.7Ghz gave: 55fps

So I guess the game is not optimized for quadcore then and a stronger cpu gives more fps.

Yep you got it