B^gD^ddy~CF~
Limp Gawd
- Joined
- Nov 5, 2006
- Messages
- 172
I was aware that there were issues encountered with 7970 Crossfire scaling in Battlefield 3 @ 2560x1600. However I encountered a scaling issue at 1920x1200. I decided to dig a little bit and do some proper testing. If you purchase anything here give something back to the [H] by using their commissioned buying links. I don't know about you but I base most of my buying decisions on what I read here.
Rig Setup
Case
Thermaltake Armor Series VA8003BWS Black Full Tower Case w/ 25CM Fan
Cooling
Noctua NH-U12P SE1366 120mm SSO CPU Cooler
3 x Thermaltake DuraMax 12 AF0060 120mm Case Fan
4 x Thermaltake TT-9025A-2B 90mm Case Fan
Motherboard
GIGABYTE GA-EX58-UD5
Processor
Intel Core i7-920 BX80601920
Memory
2 x OCZ Platinum 6GB (3 x 2GB) 240-Pin DDR3 SDRAM DDR3 1600 OCZ3P1600LV6GK
Storage
Pioneer Black Blu-Ray Burner BDR-203BKS
PLEXTOR 24X DVD Writer Black PX-880SA-26
Intel X25-M SSDSA2MH080G1 SSD
6 x WD Caviar Black WD1001FALS 1TB
WD Caviar Black WD2001FASS 2TB
Power
BFG Tech EX-1200 BFGR1200WEXPSU
GPU
XFX FX-797A-TNFC Radeon HD 7970
SAPPHIRE 21197-00-40G Radeon HD 7970
Displays
BenQ G2400WD
Lenovo ThinkVision L2440x
SAMSUNG 2443BW








As you can see that's a lot of stuff in one computer. Thats why I added more fans. The 7970's were pushing the system temperature over 69 °C when using just the fans that came with the case! The topmost 7970 was reaching 96 °C.

Anyway lets get down to bussiness.
Sofware Used
Battlefield 3 Latest version(don't know where to find the version # but executable says 1.0)
MSI AfterBurner 2.2.0 Beta 10
Catalyst 7.14.10.0890 11.12
HWiNFO64 3.91-1490 Beta
FRAPS 3.4.7
This first test was run at 1920x1200 with the settings shown in the first image below. This was a destroy the MCOM station game on a full or nearly full 64 player server. The map was Grand Bazaar. Afterburner is set to take readings every 5 seconds. HWiNFO64 does the same. GPU TEMPS, Usage, and Memory Usage were gathered with AfterBurner. System and CPU temperature Diode readings where gathered with HWiNFO64. Frames Per Second, and Min/Max/Avg Frames Per Second were gathered with FRAPS
1.)
2.)
3.)
4.)
5.)
6.)
7.)
8.)
1.) Looking at the first picture above you can see that everything is maxed out except for Antialiasing Deferred which is at 2x MSAA and Motion Blur is off because I don't like it
2.) In the second picture you can see it stays pretty close to 60 FPS for the most part except for the end where it goes to the end of round screen and 140 FPS. No drastic yo-yo effect going on which is good.
3.) In the third picture you can see that my average FPS was indeed above 60 FPS with a low of 42
4.) In the fourth picture it starts to get interesting. You can see that one of the 7970's stays at 0% utilization for almost the whole game. I'd say Crossfire scaling is broken here.
5.) In the fifth picture things are kind of weird. It makes sense that in crossfire you'd only have one VRAM usage measurement. After all both gpu's memory should have the same contents. Where it gets a little funny is the max amount of VRAM memory used shows 3166...as far as I know the 7970 only has 3072 MB of VRAM. This may be a bug in afterburner or maybe VRAM usage is measured in a different way than I think it is
6.) The sixth picture shows that my added cooling is doing the job the GPU being utilized doesn't get above 82 °C.
7.) Again the added cooling is doing it's job the System Temperature diodge doesn't go above 58 °C in extended usage
8.) The CPU stays pretty cool as well not above 56 °C.
This second test was run at 5292x1050 with the same settings as shown for the 1920 test. This was a destroy the MCOM station game on the same server as the 1920 test. Server was a 64 player server that was full or nearly full. The map was Operation Metro. It might be noted that there were a lot of people trying to cheat and getting booted in this particular game. The tickets got messed up my team ended up having 236 tickets for each set of mcoms which is why the game lasted so long. Information was collected in the same way as the 1920 test.
1.)
2.)
3.)
4.)
5.)
6.)
7.)
1.) My FPS were definately more of a mixed bag at this resolution. It wasn't horrible but it wasn't silky smooth either. I spend quite a bit more time below that 60 FPS mark on this one.
2.) Hmm the low is lower the high is lower but the average is higher. The second half of the game definately helped out here as most of that part of the game was spent indoors.
3.) Now this is more like it both GPU's are almost pegged so I don't have an expensive graphics card sitting idle.
4.) Look at that max VRAM usage number 4092....
5.) Both GPU's are being used and the GPU temps start to show it. 91 °C for GPU1 and 83 °C for GPU2. Makes me wish my motherboard was laid with more space in between the PCIE 16x slots.
6.) The GPU's aren't the only thing to heat up either the system diode gets up to a toasty 61 °C.
7.) The CPU Diode gets up to 61 °C as well.
Well there you have it. It's not even close to scientific but it's a lot better than "it felt faster". I have a whole new respect for these guys here at the [H] who have to churn out the numbers on this kind of stuff. It appears to me that Crossfire is indeed broken for 7970's @ 1920x1200. I do have a few caveats though. I brought the Sapphire first and the XFX second. When the XFX came in I ended up putting the Sapphire in the bottom slot. That and the fact that I have two 7970's from different OEMs might have something to do with my problem here. They should both be reference 7970's though so it shouldn't make a difference. It also shouldn't matter which one is in which slot but I figured I'd mention it.
Rig Setup
Case
Thermaltake Armor Series VA8003BWS Black Full Tower Case w/ 25CM Fan
Cooling
Noctua NH-U12P SE1366 120mm SSO CPU Cooler
3 x Thermaltake DuraMax 12 AF0060 120mm Case Fan
4 x Thermaltake TT-9025A-2B 90mm Case Fan
Motherboard
GIGABYTE GA-EX58-UD5
Processor
Intel Core i7-920 BX80601920
Memory
2 x OCZ Platinum 6GB (3 x 2GB) 240-Pin DDR3 SDRAM DDR3 1600 OCZ3P1600LV6GK
Storage
Pioneer Black Blu-Ray Burner BDR-203BKS
PLEXTOR 24X DVD Writer Black PX-880SA-26
Intel X25-M SSDSA2MH080G1 SSD
6 x WD Caviar Black WD1001FALS 1TB
WD Caviar Black WD2001FASS 2TB
Power
BFG Tech EX-1200 BFGR1200WEXPSU
GPU
XFX FX-797A-TNFC Radeon HD 7970
SAPPHIRE 21197-00-40G Radeon HD 7970
Displays
BenQ G2400WD
Lenovo ThinkVision L2440x
SAMSUNG 2443BW








As you can see that's a lot of stuff in one computer. Thats why I added more fans. The 7970's were pushing the system temperature over 69 °C when using just the fans that came with the case! The topmost 7970 was reaching 96 °C.

Anyway lets get down to bussiness.
Sofware Used
Battlefield 3 Latest version(don't know where to find the version # but executable says 1.0)
MSI AfterBurner 2.2.0 Beta 10
Catalyst 7.14.10.0890 11.12
HWiNFO64 3.91-1490 Beta
FRAPS 3.4.7
This first test was run at 1920x1200 with the settings shown in the first image below. This was a destroy the MCOM station game on a full or nearly full 64 player server. The map was Grand Bazaar. Afterburner is set to take readings every 5 seconds. HWiNFO64 does the same. GPU TEMPS, Usage, and Memory Usage were gathered with AfterBurner. System and CPU temperature Diode readings where gathered with HWiNFO64. Frames Per Second, and Min/Max/Avg Frames Per Second were gathered with FRAPS
1.)
2.)
3.)
4.)
5.)
6.)
7.)
8.)
1.) Looking at the first picture above you can see that everything is maxed out except for Antialiasing Deferred which is at 2x MSAA and Motion Blur is off because I don't like it
2.) In the second picture you can see it stays pretty close to 60 FPS for the most part except for the end where it goes to the end of round screen and 140 FPS. No drastic yo-yo effect going on which is good.
3.) In the third picture you can see that my average FPS was indeed above 60 FPS with a low of 42
4.) In the fourth picture it starts to get interesting. You can see that one of the 7970's stays at 0% utilization for almost the whole game. I'd say Crossfire scaling is broken here.
5.) In the fifth picture things are kind of weird. It makes sense that in crossfire you'd only have one VRAM usage measurement. After all both gpu's memory should have the same contents. Where it gets a little funny is the max amount of VRAM memory used shows 3166...as far as I know the 7970 only has 3072 MB of VRAM. This may be a bug in afterburner or maybe VRAM usage is measured in a different way than I think it is
6.) The sixth picture shows that my added cooling is doing the job the GPU being utilized doesn't get above 82 °C.
7.) Again the added cooling is doing it's job the System Temperature diodge doesn't go above 58 °C in extended usage
8.) The CPU stays pretty cool as well not above 56 °C.
This second test was run at 5292x1050 with the same settings as shown for the 1920 test. This was a destroy the MCOM station game on the same server as the 1920 test. Server was a 64 player server that was full or nearly full. The map was Operation Metro. It might be noted that there were a lot of people trying to cheat and getting booted in this particular game. The tickets got messed up my team ended up having 236 tickets for each set of mcoms which is why the game lasted so long. Information was collected in the same way as the 1920 test.
1.)
2.)
3.)
4.)
5.)
6.)
7.)
1.) My FPS were definately more of a mixed bag at this resolution. It wasn't horrible but it wasn't silky smooth either. I spend quite a bit more time below that 60 FPS mark on this one.
2.) Hmm the low is lower the high is lower but the average is higher. The second half of the game definately helped out here as most of that part of the game was spent indoors.
3.) Now this is more like it both GPU's are almost pegged so I don't have an expensive graphics card sitting idle.
4.) Look at that max VRAM usage number 4092....
5.) Both GPU's are being used and the GPU temps start to show it. 91 °C for GPU1 and 83 °C for GPU2. Makes me wish my motherboard was laid with more space in between the PCIE 16x slots.
6.) The GPU's aren't the only thing to heat up either the system diode gets up to a toasty 61 °C.
7.) The CPU Diode gets up to 61 °C as well.
Well there you have it. It's not even close to scientific but it's a lot better than "it felt faster". I have a whole new respect for these guys here at the [H] who have to churn out the numbers on this kind of stuff. It appears to me that Crossfire is indeed broken for 7970's @ 1920x1200. I do have a few caveats though. I brought the Sapphire first and the XFX second. When the XFX came in I ended up putting the Sapphire in the bottom slot. That and the fact that I have two 7970's from different OEMs might have something to do with my problem here. They should both be reference 7970's though so it shouldn't make a difference. It also shouldn't matter which one is in which slot but I figured I'd mention it.
Last edited:



