Why frames still drop in crossfire mode?

[jedirobus]

I was wondering if you could help me out here. I'm running 3 GPUs, technically, I've been playing Star Wars the Old republic. It didn't initially have X Fire support in the 11 Cat drivers so I got 12.1. It started working ok without the visuals seemingly act up and spit me out, however, I'm running a Radeon HD 5970 and an HD 5830 with X fire enabled. Am I doing something wrong because...in some areas I'll still notice really WEAK frame rates like the open terrain areas and so on. I don't know if this is normal because there is technically, alot of stuff on screen, but when I go to check performance none of the GPU activity seems to cross more then 30 perc on all GPUs at most. Now, in the big areas I've seen my frames go down to around 30, maybe less for a bit.

Am I doing it wrong? Do I need to go to "Default mode" I had heard I should run it in "1x1 optimized" but no matter which I choose my frames still tank in certain places. In confined areas I fly though...so I'm just curious if this is normal or what and if so...what was the point of even getting so many GPUs if this is happening. Any feedback would be appreciated.

My specs

Win 7 64bit
AMD Phenom II X6 @ 4.1ghz
8 GB G.Skill DDR3 1600 RAM
Radeon HD 5970 2GB
Radeon HD 5830 1GB in XFire mode
80 GB 10k RPM boot drive
150GB 7200RPM 64MB Cache game drive

 

[captjack5169]

I am speaking from personal experience so bare with me. I would pull the 5830 its not helping your situation. Check the performance once you have done that. What size is the monitor and what resolution are you using? SLI and Crossfire, were really designed to run like cards in unison. You can run off sized cards, you just shouldn't. Let me know about the screen and resolution.

 

[simon12]

When you add the 5830 it slows all the cards to the speed of a 5830 and as 3 card crossfire does not scale that well I think 1 5970 out performs 3 5830s. I could be wrong but thats what I think.

 

[jedirobus]

Well, right now I have it connected to my TV via HDMI but my max resolution is still only 1920x1080 like my monitor and my TV is 40 inches. So, you're saying that I'm actually HAMPERING my performance by running it in Crossfire? I set all the frequencies across video memory and processing to be identical thinking that would help but it hasn't changed much. I was always of the impression things like this existed to boost performance for games like Battlefield 3, which I also noticed little change but there was a slight boost and in some areas of old republic it feels like it might be worse? I'd hate to have to think I spent that money for nothing

 

[jedirobus]

It's 2 cards but the 5970 is a dual GPU card, so in essence its 3 GPUs. 2 physical cards

 

[captjack5169]

Yeah I would lose the lesser card and see what your performance is there. Run a bench tester after you make the change like the heaven tester or furmark, or even 3d mark. If your getting crappy performance there your issues are in the card. Personally I have been using one physical card for a while now. Most games are not making adequate use of sli/crossfire anyway. Yeh I think its screwing your perfomance. Remove it and bench test it both ways to find out.

 

[simon12]

The 5970 is recommended to be crossfire with another 5970 or a 5870. It is possible to crossfire it with a 5850 or 5830 but not recommended most likely due to no performance gain or performance loss.

 

[wolfram23]

I think in the case of SWTOR, the problem is lack of CF support. If 1x1 optimized isn't working well try AFR. They are different and sometimes one or the other will give much worse performance. Generally it's ideal to use Default, but if the game doesn't have a CF profile out yet then try AFR and 1x1 and see which gives best results.

Anyway, just so you know, the 5970 uses "full" cores comparable to the 5870. By pairing it with a 5830, you're significantly hampering those two cores as the 5830 has 1120 stream processors, while each 5970 core has 1600.

So while it doesn't disable the SPUs, performance wise you're going from 3200 (5970) to 3360 (3x 5830). Sort of.

Anyway, there's a good chance you have a CPU bottleneck in certain cases. The X6 aren't the best gaming CPUs. And what motherboard are you using?

Also, I wrote instructions on installing a new GPU such as for CF:
http://wolframpc.blogspot.com/2012/01/how-to-install-crossfire-sli-or-new-gpu.html

 

[captjack5169]

Yeh, this above and a little googling I just did. I would say your definate problem is the lesser card being CF on the system. Its hurting performance and making things work harder to perform. I would pull the 5830 out and bench mark the crap out of the 5970. I think you ll be impressed.

 

[jedirobus]

I think in the case of SWTOR, the problem is lack of CF support. If 1x1 optimized isn't working well try AFR. They are different and sometimes one or the other will give much worse performance. Generally it's ideal to use Default, but if the game doesn't have a CF profile out yet then try AFR and 1x1 and see which gives best results.

Anyway, just so you know, the 5970 uses "full" cores comparable to the 5870. By pairing it with a 5830, you're significantly hampering those two cores as the 5830 has 1120 stream processors, while each 5970 core has 1600.

So while it doesn't disable the SPUs, performance wise you're going from 3200 (5970) to 3360 (3x 5830). Sort of.

Anyway, there's a good chance you have a CPU bottleneck in certain cases. The X6 aren't the best gaming CPUs. And what motherboard are you using?

Also, I wrote instructions on installing a new GPU such as for CF:
http://wolframpc.blogspot.com/2012/01/how-to-install-crossfire-sli-or-new-gpu.html

I'm using an MSI 890FXA-GD70 and I know they're not the best gaming CPUs out there but I would say they aren't the worst either and, at the time, I didn't want to spend potentially double on an intel system in some areas. That's also why I overclocked it to 4.1 ghz from 3.2 to boost the performance some.

I just ran Heaven benchmark v2.1 with and without Xfire on the 5830.

Without Xfire I got a score of (1247) FPS 49.5 avg, Min: 20.1 Max: 90.5
WITH Xfire I got a score of (1468) FPS 58.3 avg, Min: 14.1 Max: 111.7

This was with Trillinear filter on, Tessellation on Moderate, Anisotropy 16x, 8x AA and 1920x1080 resolution on both tests.

So, is it just the game not being really compatible yet or a non existing profile to go with it since I came out on top with the benchmark or could those tests be off?

 

[captjack5169]

Play the game without cross fire on, see if it makes the difference you are looking for.

 

[wolfram23]

I think, generally, it's an issue with SWTOR. Try running just the 5970 with 12.1 drivers and 1X1 optimized and see if that works ok until AMD get a proper driver out for it.

 

[Earnie]

X6 is a great cpu for gaming!games like BF3 and others love more cores,Dont know where your info comes from,but its flat out wrong.

Junk the 5830 and just run the 5970

do not use a registry cleaner,your just asking for trouble.

 

[captjack5169]

X6 is a great cpu for gaming!games like BF3 and others love more cores,Dont know where your info comes from,but its flat out wrong.

Junk the 5830 and just run the 5970

do not use a registry cleaner,your just asking for trouble.

Ease up on the Fan boyness for a minute, The X6 is great process until you start comparing it to Intel current cpus. However I have built several gamers with the X6 they play just fine. The only registry cleaner I would feel safe with is CCleaner.

 

[wolfram23]

Oh, sorry, I forgot to take my crazy pills this morning!

http://www.anandtech.com/show/3674/amds-sixcore-phenom-ii-x6-1090t-1055t-reviewed/9

Anyway, I'm not saying it's bad, just not ideal. Some games don't like more cores they like fewer, faster ones. Still at 4.1ghz it should be adequate, but I don't know how SWTOR performs so it's possible that it isn't interested in 6 cores.

 

[wolfram23]

http://www.tomshardware.com/reviews/star-wars-gaming-tests-review,3087-8.html

Ok so there shouldn't be any CPU issues. So must be driver/GPU related.

OP: You installed the two GPUs in slots 1 and 5, right?

 

[Earnie]

Ease up on the Fan boyness for a minute, The X6 is great process until you start comparing it to Intel current cpus. However I have built several gamers with the X6 they play just fine. The only registry cleaner I would feel safe with is CCleaner.

fanboyness?....always wondered why people make idiotic comments like this,I've used both!
It has nothing to do with AMD or intel,its about cores,believe what you want.

 

[alrobichaud]

I think, generally, it's an issue with SWTOR. Try running just the 5970 with 12.1 drivers and 1X1 optimized and see if that works ok until AMD get a proper driver out for it.

It does sound like a problem with SWTOR if you are only seeing 30% gpu usage while playing. Though your configuration works, it is not a recommded configuration by AMD so you have to beleive that they do not recommend it for a reason.

IF you compare the specs, the 5970 has more than double the stream processors, textuire units, etc of the 5830. I imagine the 5970 is seriously getting bottlenecked by trying to run in unison with the 5830. Simply adjusting the gpu and memory speeds will not help this situation. I think your heaven benchmark tests are fairly accurate consdering what you are crossfiring. They do have the 5870 listed as a recommended configuration. It exaclty matches the architecture each of the gpu's in the 5970.

*edit*

Here is the link to the AMD chart. Sorry for the poor quality pic.

http://sites.amd.com/us/game/technology/Pages/crossfirex.aspx

 

[jedirobus]

So, what you're saying is...It is correct in showing that I'm getting a boost and it's the game's fault, currently and the current drivers most likely? The reason I'm getting such a slight boost is because the architecture, even though is in the same generation, is too different. I mean it was a 100 dollar investment, basically to try and boost performance and the Heaven benchmark shows that it goes up but not by much it seems so I guess thats why I'm noticing little or nothing in SWTOR and since it's still been out only a month AND with new drivers its probably still not really good for it yet. Either way I probably should have just paid the extra 70 bucks or so maybe. Might be worth trying to resale and just get a 5870 then If I want more a performance boost due to compatibility on the cards?

 

[wolfram23]

Pretty much, yeah.

Have you tried SWTOR with just the 5970?

 

[alrobichaud]

So, what you're saying is...It is correct in showing that I'm getting a boost and it's the game's fault, currently and the current drivers most likely? The reason I'm getting such a slight boost is because the architecture, even though is in the same generation, is too different. I mean it was a 100 dollar investment, basically to try and boost performance and the Heaven benchmark shows that it goes up but not by much it seems so I guess thats why I'm noticing little or nothing in SWTOR and since it's still been out only a month AND with new drivers its probably still not really good for it yet. Either way I probably should have just paid the extra 70 bucks or so maybe. Might be worth trying to resale and just get a 5870 then If I want more a performance boost due to compatibility on the cards?

So forgetting about SWTOR and using heaven 2.1 as a guide, you are looking at roughly an 18% increase going from 49.5FPS to 58.3 FPS. If you did upgrade a 5870 and if it did scale perfectly you would be looking at taking 49.5 FPS and adding 25 FPS which would be half of the 5970 for a total of 75 FPS. I do not know how well those cards scale and I am sure it will change depending on the game you are playing but I think 90% would be a reasonable number to aim for. Based on 90% add 49.5 to 22.5 for a total of 72FPS using the heaven benchmark. It won't be a lot more but it will still be better than using the 5830. So the question now becomes is it worth it to you to resell the 5830 and buy a 5870 to gain another 14FPS.

 

[jedirobus]

Good question...I dunno if it is worth it then or not. What do you recommend then?

Also I have tried it with just the 5970 and the performance is about the same perhaps if not better when I tell it to use the 5970 in 1x1 optimize instead of X fire 1x1 optimize with the 2 cards. Maybe its somehow skipping over the second R870 chip and going to the 5830 or something. It's probably game related in someway since its still relatively new. I'm kind of at an impass on what to do then to effectively boost efficiency, especially now that I already bought a 5830 the question does become do I resell and buy a 5870 for what would only be 14 FPS, at best maybe not even that?

 

[Haserath]

I would not recommend a 5970 be paired with a 5830.

A 5970 is essentially two 5870's, each with 1600 stream processors/80 Texture units/32 ROPs. The 5830 is a card with 1120 stream processors/56 Texture units/16 ROPs. This match up, along with Alternating Frame rates, the technique used to produce frames in Crossfire setups, would most likely produce worse results than better results, because the two 5870's would be very uneven with the 5830.

I just ran Heaven benchmark v2.1 with and without Xfire on the 5830.

Without Xfire I got a score of (1247) FPS 49.5 avg, Min: 20.1 Max: 90.5
WITH Xfire I got a score of (1468) FPS 58.3 avg, Min: 14.1 Max: 111.7

Notice the minimum framerates between the systems? The 5830 helps produce more raw frames(average), but the consistency is worse(minimum rates). Consistent frames are more important than high average frame-rates.

What are your settings for SWTOR? Do you know how much Vram it uses when you have those low frame-rates?

 

[jedirobus]

No, I can't say I know how much VRAM it uses. I don't really know how to check how much VRAM a game is using while it's running. I can only tell how much memory it's using via task manager. Unless that's what you mean

 

[wolfram23]

GPUZ and MSI Afterburner should report VRAM used.