Can devs enlighten regarding dual GPU

[SolitarioSoldat]

Devs whats the problem now with the dual ATI's cards,new CAP1 was released specially for 11.9 drivers but seems that performance got very minor boost and still seeing sometimes only 25-28% usage on both GPU's at the same time....

I would assume its not the new drivers at fault but the game itself,can you guys be honest whats going on?

 

[VilMik]

I know very little about... well... everything... so dont trust me too much, but i think its mostly about the drivers, especially for crossfire/sli.
11.9 have not been tweaked at all for RO2.

I have read somewhere that ati is tweaking for future drivers.
So maybe with luck their 11.10 version, or 11.11.

Have you tried atis 11.10 beta drivers? the beta version have no RO2 optimization, but they gave me some extra fps. (single 5770 card)

Remember i could be wrong, and most likely are

 

[Richey79]

Game's CPU intensive, so you're not going to see great improvements using CrossFire. With some CPU intensive games, you actually get worse performance than with one GPU.

The devs have said they hope to be able to off-load more work to the GPU in the future, but it's a complex task changing the core code in this fashion, as you'd imagine.

 

[SolitarioSoldat]

Game's CPU intensive, so you're not going to see great improvements using CrossFire. With some CPU intensive games, you actually get worse performance than with one GPU.

The devs have said they hope to be able to off-load more work to the GPU in the future, but it's a complex task changing the core code in this fashion, as you'd imagine.

So you say is CPU intensive? My specs are in my sig so here is what get while gaming:

CPU load: 35%
GPU1 load: 25-40%
GPU2 load: 25-40%

How is 35% CPU load = game being CPU intensive???

Dont want to name which game but here is what I get there,really nice hardware usage and FPS also high:

CPU load: 85%
GPU1 load: 95-99%
GPU2 load: 95-99%

Call me stupid but the one above is what I call properly optimised game...

ATI 11.9 CAP1 came with RO2 Crossfire profile,we cant blame all on drivers,its obvious its not a driver issue,just saying.

 

[UncleDrax]

How is 35% CPU load = game being CPU intensive???

How many CPUs, Cores do you have, and do you have HT?

 

[abjectblitz]

How many CPUs, Cores do you have, and do you have HT?

I get EXACT same result as him regarding usage.

I'm on a 5970 and Q9650 @ 4.2ghz

 

[blkmgc]

Dont want to name which game but here is what I get there,really nice hardware usage and FPS also high:

CPU load: 85%
GPU1 load: 95-99%
GPU2 load: 95-99%

Call me stupid but the one above is what I call properly optimised game...

Probably a different game engine so....

 

[Kasspa]

How many CPUs, Cores do you have, and do you have HT?

Don't mean to sound like an ***, but can you seriously not read? He said his specs were in his signature... hes got an AMD 6 core processor.

 

[wazups2x]

My brothers 2 4850's run great after the latest driver (11.9) release and cap1 (Red Orchestra 2 crossifre fix) . Now he can actually run it on Ultra.

 

[SolitarioSoldat]

How many CPUs, Cores do you have, and do you have HT?

Its the AMD 1090T x6 @ 4.0ghz

My brothers 2 4850's run great after the latest driver (11.9) release and cap1 (Red Orchestra 2 crossifre fix) . Now he can actually run it on Ultra.

Crossfire for the part that game is more stable with new CAP1 its true,but not much of an advance over using single GPU thats been stressed up to 70% usage...

Something aint right here,dont want to point fingers towards the new drivers nor towards the devs,new drivers made things better,but the GPU's are still just jerking around barelly breaking a sweat.

As for the CPU is also barelly used,at 35% max and most of the time between 28-30%,I can assume 3 cores are being (Not Fully) used at the most,not CPU intensive at all.

 

[Dirtybandit]

Been waiting a while for a fix, the game is unplayable for me. constant stuttering and huge fps drops ;(

 

[KompanyCommands]

I have 2x 5870's with slightly poor performance with an I7 Quad.


However on the beta a patch seemed to smooth everything out..


Lets hope they do the same on the final release and my laptop is the AW M17X R2 2010 edition.

 

[nu3gawhat]

The problem isn't limited to ATI, I have a GTX295.

I have way worse frames with SLi enabled than disabled.

Haven't played the beta patch, so don't know if there's been any improvement there.

 

[Geamr]

I'm almost glad my gtx295 pooped out a few months ago. Got a GTX570, and i won't buy a dual GPU card again any time soon. Only had problems with that thing.

 

[UncleDrax]

Its the AMD 1090T x6 @ 4.0ghz
...
As for the CPU is also barelly used,at 35% max and most of the time between 28-30%,I can assume 3 cores are being (Not Fully) used at the most,not CPU intensive at all.

You might be assuming incorrectly btw. (I'm just saying).
So it's 6 cores? and you're using 35%? that for all processes, or just RO2game.exe?
My point before about you just linking % of total CPU used was to point out you can't just do that sorta thing anymore since almost every machine is multi-CPU|core|hyperthreaded. 15% of a 8-core system is 1 whole core being slammed.. for example. And you can still have serious problems if the process that is using that one core is bottlenecking, driving load up (load being the *nix version definition here)

I don't know for certain since I haven't any any issues personally with ro2, I'm just wondering if it is actually a CPU bottle neck, and if it is, if you situation is one in which you are having some sorta threading issue. I'm sure RO2's main() thread will almost always be allocated to a single core (I would be as bold to say that it's true of almost any main() type software with a super-parent handler.. game engines are prime examples of that)

 

[ComeAndSee]

Strange, I have a Sapphire Toxic 2 GB 6950 flashed to 6970 and I'm not experiencing any of the issues you guys are having.

Your guys video cards may just not be good enough for the game on high settings. The 5870 and GTX 295 are mid range cards. I went from a 5950 x2 to my 6950 and it was a massive performance increase.

 

[SolitarioSoldat]

Strange, I have a Sapphire Toxic 2 GB 6950 flashed to 6970 and I'm not experiencing any of the issues you guys are having.

Your guys video cards may just not be good enough for the game on high settings. The 5870 and GTX 295 are mid range cards. I went from a 5950 x2 to my 6950 and it was a massive performance increase.

I can tell your knowledge is on the low side of hardware,here is why.

your precious 6950 is maeby at most 10% stronger than the 5870,we are talking with people having issues with 2 cards,the dude with 2x5870 will wipe the floor with your single 6950 iven if you flush the bios to 6970 speeds.

I would of say CF another 6950 and you will see that the bump you get with CAP1 makes no sence since card are running to max at 50% and thats it,most of the time cards are stressed at 40%...While other games stress both GPU's at 99%!!!!!!!!!!!! Seems RO2 have problem with taking advantage from our single and multi GPU setup.

Before CAP1 for 11.9 CF setup was actually giving you 20% less than a single setup... Download MSI afterburner and see how much your game utilizes from your card,probably not higher than 60-70%.

 

[SolitarioSoldat]

You might be assuming incorrectly btw. (I'm just saying).
So it's 6 cores? and you're using 35%? that for all processes, or just RO2game.exe?
My point before about you just linking % of total CPU used was to point out you can't just do that sorta thing anymore since almost every machine is multi-CPU|core|hyperthreaded. 15% of a 8-core system is 1 whole core being slammed.. for example. And you can still have serious problems if the process that is using that one core is bottlenecking, driving load up (load being the *nix version definition here)

I don't know for certain since I haven't any any issues personally with ro2, I'm just wondering if it is actually a CPU bottle neck, and if it is, if you situation is one in which you are having some sorta threading issue. I'm sure RO2's main() thread will almost always be allocated to a single core (I would be as bold to say that it's true of almost any main() type software with a super-parent handler.. game engines are prime examples of that)

Its the 35% at the most,90% of the time when in game is at 28-30% yes with all processes not just the game....

I know something is not right here,CPU and GPU are poorly used by RO2.

 

[Yoshiro]

Unreal Engine 3 uses 2 main threads. The Game Thread and the Rendering Thread. Both try to split to their own core, and other threads if possible will spread out to cores 3 and 4 if they are found. If not they slot into cores 1 and 2. Each main thread is a single core process.

The problem we are/were running into was with the rendering thread bottle-necking on the amount of draw calls on the lower end CPU's. We have put our emphasis into optimizing that. As to how the engine makes use of cores/gpu beyond that we are in line with the engine norm. As always we'll look for areas to improve, but if you grab the September 10/11 UDK demos and run the the demos in both you will see similar usage of the cores, with gpu going between 60 to 90 % usage depending on the scene.

 

[ComeAndSee]

I can tell your knowledge is on the low side of hardware,here is why.

your precious 6950 is maeby at most 10% stronger than the 5870,we are talking with people having issues with 2 cards,the dude with 2x5870 will wipe the floor with your single 6950 iven if you flush the bios to 6970 speeds.

I would of say CF another 6950 and you will see that the bump you get with CAP1 makes no sence since card are running to max at 50% and thats it,most of the time cards are stressed at 40%...While other games stress both GPU's at 99%!!!!!!!!!!!! Seems RO2 have problem with taking advantage from our single and multi GPU setup.

Before CAP1 for 11.9 CF setup was actually giving you 20% less than a single setup... Download MSI afterburner and see how much your game utilizes from your card,probably not higher than 60-70%.

More like 30% better and 1/3rd the price single GPU vs. >_>