6970 Crossfire under performing?

[jon_the_faun]

I recently built a new gaming rig, and I've been ripping my hair out ever since. Here are my specs:


AMD Phenom II X6 1100T (with stock cooling)
ASUS Crosshair IV Formula AM3 AMD 890FX SATA 6Gb/s USB 3.0 ATX AMD Motherboard
2 SAPPHIRE 100311SR Radeon HD 6970s in Crossfire
COOLER MASTER HAF 932 Advanced
G.SKILL Ripjaws Series 8GB (4 x 2GB) 240-Pin DDR3 SDRAM DDR3 1600
SAMSUNG Spinpoint F3 HD103SJ 1TB 7200 RPM 32MB Cache SATA 3.0Gb/s 3.5" Internal Hard Drive -Bare Drive
Ultra X4 1050-Watt Modular Power Supply


Being my first gaming computer, I expected to run into some errors. I'm not an ace when it comes to the details, but I do know more than the average user (hardware functions/plugs/BIOS). With some guidance, perhaps I can get this dilemma solved:

I've spent hours on the forums here at Tom's Hardware and on other forums elsewhere that were suggested by Google. I've read all the benchmarks for my cards, and downloaded all the little tools off CNET (CPU-Z, GPU-Z....which reads that I do, indeed, have Crossfire Enabled). I have downloaded the latest drivers (11.6) from AMD. And yet, when I run FRAPS...I'm not getting the same performance reads.

For example:

Just Cause (MAX Settings, 1680 x 1050) --- 60,61,60,59 Maximum fps.......55,54,45,40 Minimum fps
Witcher 2 (High Settings, 1680 x 1050...didn't even both with the ubersampling) --- 50,49,48,55, Maximum fps......45,44,39 Minimum fps) I downloaded the hotfix that enables Crossfire Scaling

However, BFBC2 runs at a solid 60 FPS on max settings...which makes me happy inside. Still...I built this computer for the sole purpose of BF3 and Skyrim...with my inconsistent results...I don't think I'll be having the best time blowing stuff up and slaying dragons.

What I don't get is how people get 130 FPS using these cards in Crossfire. Even on BFBC...it stutters between 60,61,59 and never exceeds those numbers.

Somebody please help. Before I start cooking my computer parts for food because they've left me broke and unhappy. 🙁

 

[ubercake]

It seems like you're still running in single-card mode somehow:

http://www.guru3d.com/article/radeon-6950-6970-review/22

A 6950 crossfired will give you better performance in BFBC2:

http://www.guru3d.com/article/radeon-hd-6950-crossfirex-review/12

Did you originally start with one card and then recently add the second or have you always had two 6970 cards?

 

[jon_the_faun]

Yeah, that's what I thought as well.

I originally started with two 6870s...then decided - after much deliberation - to upgrade to the 6970s. But I should also note that after I play a game, GPU-Z reads both cards at 65-70 C. Does that necessarily mean that both are being used?

 

[bystander]

Given that almost all your games are maxing at or very close to 60 FPS, I think you have v-sync forced on. You may also have Morphological AA on, or some other AA on. Check the AA mode as well. Supersampling AA is a killer on performance.

 

[ubercake]

You can create a log of GPU load using GPU-z as well. Setup GPU-Z to log info for card 2 and look at the log file after running the game. You're looking for consistently low percentages in the load for your second card.

Also, like bystander said, if you're forcing AA on through Catalyst or in the game settings, this really can reduce the frame rates, but the trade-off is the the better quality. But like the guru3D review shows, even at 8xMSAA, the 6950 crossfire setup will give you 96fps on average in BFBC2.

 

[bystander]

You can create a log of GPU load using GPU-z as well. Setup GPU-Z to log info for card 2 and look at the log file after running the game. You're looking for consistently low percentages in the load for your second card.

Also, like bystander said, if you're forcing AA on through Catalyst or in the game settings, this really can reduce the frame rates, but the trade-off is the the better quality. But like the guru3D review shows, even at 8xMSAA, the 6950 crossfire setup will give you 96fps on average in BFBC2.

I'm pretty certain he has v-sync on in this game at the least. He has a solid 60 FPS, which is the refresh rate of most monitors. Games almost never have solid FPS at a single number without v-sync.

 

[ubercake]

I'm pretty certain he has v-sync on in this game at the least. He has a solid 60 FPS, which is the refresh rate of most monitors. Games almost never have solid FPS at a single number without v-sync.

I know what you're saying, though he's showing a good range of FPS...

From the original post:
...Just Cause (MAX Settings, 1680 x 1050) --- 60,61,60,59 Maximum fps.......55,54,45,40 Minimum fps
Witcher 2 (High Settings, 1680 x 1050...didn't even both with the ubersampling) --- 50,49,48,55, Maximum fps......45,44,39 Minimum fps) I downloaded the hotfix that enables Crossfire Scaling...

Just Cause from 40 up to 61.
Wicher 2 from 39 up to 55.

The numbers for BFBC2 clearly indicate vsync, so I think it's just an in-game setting there.

 

[jon_the_faun]

Hmm. I opened up AMD VISION Engine Control Center...went to Gaming....3D Application Settings....

+ Morphological filtering is off.

+ Anti-Aliasing is on "Use Application settings"
+Anisotropic Filtering is also on "Use Application settings"
+Tessellation..."Use Application Settings"

+ Catalyst A.I - the "Texture Filtering Quality" bar is set to "Quality"

+ "Enable Surface Format Optimization" box is unchecked.

+ "Wait for vertical refresh" bar is on the 2nd line, which says, "Off, unless application specifies"

+ "Anti-Aliasing Mode" bar is on the 1st line, which says, "Multi-sample AA"

+ OpenGL Settings...Triple buffering box is unchecked.


I'll also note that the Control Center has the Crossfire box checked...but when I click Identify GPU...it only says 1. GPU-Z says otherwise....it acknowledges that I have Crossfire Enabled.

Not sure if that info helps identify the issue with my case here...but I thought I'd provide it anyhow.

Going to create a log on GPU-Z while playing Just Cause 2.

Will check forum immediately after.

Thank you guys for your fast response and help in this irritating situation.

 

[jfby]

Crossfire can be annoying... I'm looking forward to your results.

I have had so many crossfire issues just getting it to work, my next graphics solutions will probably be single card, or SLI... I keep hoping for better drivers, but to no avail.

 

[hdawood]

Make sure Vsync is disabled from in-game settings.

To verify if crossfire is enabled in a game or not, right click Catalyst Control Center icon in the taskbar, select the first graphics card that you see in the list, and a new menu would pop up, look at the bottom for Crossfire Icon and click to enable it. Once done, you should see crossfire logo on top right of your screen when running games or any other 3D application. This confirms crossfire is enabled. If there is no logo, crossfire might not be working properly.

When installing drivers, select custom option and 'check' application profiles at the top (it is unchecked by default). Also try downloading latest Catalyst Application Profiles (CAP) from AMD website.

Hope this helps!

 

[jon_the_faun]

After playing Just Cause 2 for about 30-40 minutes...the GPU Load percentages fluctuated between 20 and 45. What does that mean?

 

[jon_the_faun]

Make sure Vsync is disabled from in-game settings.

To verify if crossfire is enabled in a game or not, right click Catalyst Control Center icon in the taskbar, select the first graphics card that you see in the list, and a new menu would pop up, look at the bottom for Crossfire Icon and click to enable it. Once done, you should see crossfire logo on top right of your screen when running games or any other 3D application. This confirms crossfire is enabled. If there is no logo, crossfire might not be working properly.

When installing drivers, select custom option and 'check' application profiles at the top (it is unchecked by default). Also try downloading latest Catalyst Application Profiles (CAP) from AMD website.

Hope this helps!

Well, it appears Crossfire is running. I followed the steps you provided and the ATI Crossfire logo appeared in the top right corner of the game menu. I don't understand why its not meeting its full potential. After disabling both AA and Vysnc in the game...I still have the same FPS in Just Cause 2 (Max 58/Minimum 40) According to some benchmarks I read, with my 1680x1050 monitor, maxed out settings...these numbers are equal to one SINGLE 6970's performance.

Very frustrated.

 

[jon_the_faun]

So now...gone to my start up bar....right clicked on the Control Center...set the texture filtering quality to Performance instead of Quality...and FRAPS is giving me 50 MIN - 101 MAX in Just Cause....really sporadic stuff.

 

[bystander]

Keep in mind, the benchmarks shown are from a specific area, not all areas of the game are the same. Maybe you should run the benchmark they use.

If may be of use to try using 3dmark11 as well, so we can more accurately see how your system is performing. Or at least, see if it's performing well at all.

 

[redeemer]

Uninstall and reinstall the drivers

 

[jon_the_faun]

Keep in mind, the benchmarks shown are from a specific area, not all areas of the game are the same. Maybe you should run the benchmark they use.

If may be of use to try using 3dmark11 as well, so we can more accurately see how your system is performing. Or at least, see if it's performing well at all.

Will do. Gonna download 3dmark11 now. Will report back with results.

 

[jon_the_faun]

Bad news for me...this is my score: P8087 3DMarks

****RESULTS****

Your Score is Low Compared to Similar Systems.
There may be a problem.


PROCESSOR
******************************

Processor AMD Phenom II X6 1100T
Processor clock 3311 MHz
Physical / logical processors 1 / 6
# of cores 6



GRAPHICS CARD
******************************

Graphics Card AMD Radeon HD 6970
Vendor Advanced Micro Devices Inc.
# of cards 2
SLI / CrossFire On
Memory 3 MB
Core clock 0 MHz
Memory clock 0 MHz
Driver name AMD Radeon HD 6900 Series
Driver version 8.861.0.0
Driver status Not FM Approved


GENERAL
****************************

Operating system 64-bit Windows 7 (6.1.7601)
Motherboard ASUSTeK Computer INC. Crosshair IV Formula
Memory 8192 MB

Module 1
2048 MB G.Skill 9 @ 667 MHz

Module 2
2048 MB G.Skill 9 @ 667 MHz

Module 3
2048 MB G.Skill 9 @ 667 MHz

Module 4
2048 MB G.Skill 9 @ 667 MHz

Hard drive model1,000GB SAMSUNG HD103SJ ATA Device


DETAILED SCORES
***************************

3DMark Score P8087
Graphics Score 10176
Physics Score 5288
Combined Score 4633
Graphics Test 1 44 FPS
Graphics Test 2 54 FPS
Graphics Test 3 65 FPS
Graphics Test 4 29 FPS
Physics Test 16 FPS
CombinedTest 21 FPS

SETTINGS
***************************

Default settings used Yes
Edition Basic
Screen Width 1280
Screen Height 720
Msaa Sample Count 1
Texture Filtering Mode Trilinear
Max Af Anisotropy 1
Tessellation Detail 5
Max Tessellation Factor 10
ShadowMapSize 5
Shadow Cascade Count 4
Surface Shadow Sample Count 16
Volumetric Illumination Quality 5
Ambient Occlusion Quality 5
Depth Of Field Quality 5
Enable Window Mode Off
Enable Vertical Sync Off
Enable Triple Buffering Off
Enable Wireframe Off
Color Saturation 100%

 

[bystander]

That score is not actually low. It's about what you'd expect from a stock setup. A single card would get you about a 5k score. It could be a touch low, but it's not far below normal. I score in the 9k with my setup, which is overclocked a lot. With unlocked 6950's at 915/1375 and my CPU, which is a fair better faster than yours, I scored 10k even.

All stock setups, when compared on 3dmark, will show performing lower than average. This is because most scores stored on their system, are overclocked, many to the extreme.

 

[jon_the_faun]

That score is not actually low. It's about what you'd expect from a stock setup. A single card would get you about a 5k score. It could be a touch low, but it's not far below normal. I score in the 9k with my setup, which is overclocked a lot. With unlocked 6950's at 915/1375 and my CPU, which is a fair better faster than yours, I scored 10k even.

All stock setups, when compared on 3dmark, will show performing lower than average. This is because most scores stored on their system, are overclocked, many to the extreme.

Why are my Core and Memory Clocks at 0 under the Graphics section? And what's with the driver status "Not FM Approved"? Also, under settings, it says the screen width and height (resolution?) 1280x720. Isn't that really low, and shouldn't that mean that because of the lower resolution, its less load on the GPU's...and the peformance should be much better?

I heard that the 6970's don't overclock well, compared to the 6950s...but doesn't a 6950 overlocked = 1 slightly better than stock 6970?

 

[jon_the_faun]

Sorry for the bombardment of questions...it's just desperation. This is the first time I've ever saved up enough money for a good PC...and I just feel a tad let down. Maybe my expectations were too high. That's not an unusual thing in my life. 🙁

 

[jon_the_faun]

After spending a solid 2 months reading up and waiting for new drivers to come out...I made my decision to buy these 6970s...I mean...I wanted a single card solution, cause Crossfire has too many problems, but the 6990 was too rare, too hot, and too unreliable from the reviews I read. So I dished out the same amount of money on the next best thing: 6970s.

I only hope that future drivers will leverage the capabilities as the developers begin to further understand their cards and their relationships to upcoming games like BF3, Bioshock Infinite, and Skyrim. Because as of right now....my wallet is empty....and so are my hopes...

 

[bystander]

Why are my Core and Memory Clocks at 0 under the Graphics section? And what's with the driver status "Not FM Approved"? Also, under settings, it says the screen width and height (resolution?) 1280x720. Isn't that really low, and shouldn't that mean that because of the lower resolution, its less load on the GPU's...and the peformance should be much better?

I heard that the 6970's don't overclock well, compared to the 6950s...but doesn't a 6950 overlocked = 1 slightly better than stock 6970?

I'm not sure about why is doesn't read the core and memory clocks, obviously it's just not being read by the benchmark.

The drivers are not approved, because you have up to date drivers. 3dmark11 is always 2-3 months behind on what is up to date, and new drivers are not FM approved.

The resolution for the benchmark is always 1280x720. This is so everyone can compare more easily, as higher resolutions slow down performance.

 

[bystander]

After spending a solid 2 months reading up and waiting for new drivers to come out...I made my decision to buy these 6970s...I mean...I wanted a single card solution, cause Crossfire has too many problems, but the 6990 was too rare, too hot, and too unreliable from the reviews I read. So I dished out the same amount of money on the next best thing: 6970s.

I only hope that future drivers will leverage the capabilities as the developers begin to further understand their cards and their relationships to upcoming games like BF3, Bioshock Infinite, and Skyrim. Because as of right now....my wallet is empty....and so are my hopes...

6970's are better than a 6990. You clearly aren't aware that a 6990 is two 6970's on a single card. They are in crossfire, and they are downclocked, which is why they are slower.

 

[jon_the_faun]

No, on the contrary - I am aware of that. That's why I actually BOUGHT the 6970s.

Read that in this http://www.guru3d.com/article/radeon-hd-6990-review/

The advertising states that its THEIR BEST GRAPHICS CARD. If my 6970s stutter like a retard on coke when playing Just Cause 2 without vysnc or AA...then what AMD card out their can reach my expectations...which aren't really THAT high compared to much of the boasting I read on forums. I mean:

60 FPS...no more...no less, on a 1680x1050 monitor, while playing Just Cause 2 which came out almost a year ago. That's not too much to ask for, right? I feel robbed. And my hopes were that it was based on my own error in building of the 1600 dollar computer...but its become apparent that its actually AMD's fault.

 

[jon_the_faun]

My frustration is fueled only by the false promise of near perfection these card owners claim to have in their reviews..."YEAAAAAAAAAH DUDE! 6970s pillage and rape the Witcher 2 on Ultra!!".."Dude, 6970's are monsters. They crush souls and run every single game out their max - except Metro!!"

...that kinda stuff....

Would you not agree? I mean...what kind of performance do you get when playing Just Cause 2 or Witcher...or Crysis/Warhead/2? I'll bet you a million that they are better than my results.