Joined
·
3,534 Posts
I just got done with testing out my new MSI 5850 Twin Frozr. Sorry if the pics look bad, I only got a crappy camera phone.
-Unboxing
![]()
![]()
Box has a bunch of spec on it (see below for details)
![]()
![]()
Stuff it came with. 2x Molex to PCI-E, Afterburner software (obsolete, see below), DVI to VGA adapter, NO crossfire bridge.
-Card Closeups
![]()
![]()
![]()
![]()
![]()
-Test Setup
This test will primary consist of comparing temperatures while running benchmarks and GPU stability tests. Not so much interested in how many more FPS I could get compared to the reference card.
Test rig (my sig rig):
-AMD Phenom II x4 965BE @ 3809.5 MHz
multiplier: x19.0, bus speed: 200.5MHz
-G.Skill Ripjaws Series 2x2GB DDR3 8-8-8-24-1T 1333MHz
-MSI 790GX-G65
-CM V8 CPU cooler
-Corsair TX650
-Samsum F3 1TB and Seagate Barracuda 500GB
-CM HAF 922
fan configuration - intake: 200-front, 120-botom, 80-window. output: 200-top, 120-back
Both cards use the same Afterburner fan speed vs temp graph:
![]()
Ambient temperature sensor: Fluke 51 series, with probe positioed betweem VGA and CPU cooler
![]()
Initial ambients are about 21-23C. I opened my window after every test to get the hot air out before doing another test.
Overall setup (note the companion cube plushie)
![]()
-Interesting MSI 5850 Twin Frozr specs
All the specs are the same as any reference clock. Though I did find the core voltage to be 1.100V default as compared to my reference card at 1.087V. The fan comes preset to 30% as opposed to reference at 20%.
MSI says it has dual ultra quite PWM fans that provide 50% better thermal efficiency and direct airflow to cool the GPU. It has a large copper base coated with nickel that provides "feastest" heat dissipation (typo on the box). 8mm supper heat pipes. Ultra high-density aluminum heatsink has large area for heat dissipation. Claims 12C cooler under full load (shows a graph of the frozr vs a reference with the frozr at 76C and the reference at 88C, states "Results is based on 25C of system temperature). It has military grade componets including, highly conductive caps, solid caps, and solid state chokes.
-Cards to be compared
![]()
Sapphire 5850 reference card and MSI 5850 Twin Frozr
-Tests
Both cards are being tested with stablity and benchmarks at stock settings and overclocked to 840MHz on the core and 1150MHz on the memory at 1.112V on the core. I recorded the ambient temperature using my Fluke, and I used GPU-Z 0.4.0 and its sensor logging to record the GPU temps during th test.
Tests and settings
-FurMark 1.7.0, run for 5 minutes
Run Mode: Stability Test, checked Xtreme Burning Mode and Log GPU temperature
Window Params: Fullscreen, 1920x1080, MSAA 8X
-Unigine Heaven 2.0
API: dx11
Shaders: high
Tessellation: extreme
Anisotropy: 16
Anti-aliasing: 8x
Fullscreen, 1920x1080
-Call of Pripyat
Preset: ultra
Renderer: dx11
1920x1080
MSAA: 4x
MSAA for A-tested objects: dx10.1 style
SSAO: HDAO
SSAO Quality: ultra
checked: use dx10.1, enable tessellation, contact hardening shadows
-Far Cry 2 benchmark tool
map: small ranch
1920x1080, fullscreen, vsync
Performance: all very high
Overall Quality: ultra high
-GPUTool 0.0.0, run for 5 minutes
checked: sync clocks of multi-GPU, renderer uses fullscreen
everything else default
-Dirt2 benchmark
1920x1080, 60Hz, vsync
MSAA: 8x
everything else turned to highest setting
-FPS Results
-Heaven
Sapphire stock
![]()
Sapphire OCed
![]()
MSI stock
![]()
MSI OCed
![]()
-Call of Pripyat
Sapphire stock
![]()
Sapphire OCed
![]()
MSI stock
![]()
MSI OCed
![]()
-Far Cry 2
Sapphire stock
![]()
Sapphire OCed
![]()
MSI stock
![]()
MSI OCed
![]()
-Dirt 2
Data pulled form the XML dirt 2 produces
![]()
-Temperature Results
Only the OCed temps are in the graphs, mainly because I think they are the more important ones.
![]()
-Notes on tests
EDIT: Each card has now done each test on the same drivers, Cat 10.3 and Afterburner 1.5.1
-Result Discussion
I initially bought the MSI card due to its advertised coolness, aka 12% less the reference, and quiter fan speed. The fan did turn out to be much quiter. Up to about 60-70% fan speed, the sound level is less than that of the reference. The temperature results of the MSI are a bit unsettling to me. While I see that the graph posted on the box is for very high temps (70C+), I would think I would see even slightly better temps at lower loads. Things get even worst with OCing.
Clock for clock comparison to the Sapphire, there is no noticeable FPS gain that I can tell. I expected this, and thus really tried to focus on the temps camparision.
The Newegg advertisement says "OverVoltage Support by using latest AfterBurner 1.5.1". I have Afterburner set up correctly so I can control the voltages and such. At 1.112V, the card was fine. At 1.150V the card freaks out under load, regardless of the core and mem clock frequencies. Starting GPUTool causes the screen to go black, and eventually stops giving a signal, and the fan speed goes over 60% or so. I got through 25 stages of Heaven with 1.149V, but artifacted like crazy on the last stage. So far I am not happy.
I am in contact with MSI about these issues. Hopefully they get resolved. I may eventually flash the BIOS and try going from there. But for now, this card is not worth the $329 I spent. MSI has since emailed me back offering no help since "we do not support overclocking because it puts ware and tear on the parts. Also there is no guarantees that overclocking it will work".
-MSI With cardboard shroud test results
Added a shroud to see if I could reduce the temps by more efficiently removing the hot air out the back. Without the shroud, the air from the fans will just enter the case. The shroud is made from a cereal box and cover the whole card with holes for the fans and held on with scotch tape. Also used the tape to seal other openings, though some still exist.
![]()
Ran the same battery of tests again in the same testing conditions. Temp results:
![]()
Very slight improvements in GPU temps, but big improvement in ambient temps. This is most likely caused by less leakage into the case.
-Final thoughts
Over the course of the test, the fan noise was very minimal compared to the reference card. I have a few conclusions on this that also relate to the temperature:
1. The fan blades are significantly smaller than other cooling fans on other cards that are positioned on top of the heat sinks. Thus, there is less airflow and less noise produced by the fans.
2. Assuming #1 is right, then less airflow would mean worst temps compared to the reference. It seems to me that my tests support this.
Some how, MSI was able to produce temps of 76C on the Frozr and 88C on a reference, thus allowing them to say a 12C cooler card. This seems a little extreme to me, since none of my tests gave temps of this magnitude, though I am using a different setup and conditions (MSI doesn't state theirs).
If you are looking for a cooler card, I would say stay away and get a water block. If you want a quieter card, then this should be a consideration.
-Unboxing
Box has a bunch of spec on it (see below for details)
Stuff it came with. 2x Molex to PCI-E, Afterburner software (obsolete, see below), DVI to VGA adapter, NO crossfire bridge.
-Card Closeups
-Test Setup
This test will primary consist of comparing temperatures while running benchmarks and GPU stability tests. Not so much interested in how many more FPS I could get compared to the reference card.
Test rig (my sig rig):
-AMD Phenom II x4 965BE @ 3809.5 MHz
multiplier: x19.0, bus speed: 200.5MHz
-G.Skill Ripjaws Series 2x2GB DDR3 8-8-8-24-1T 1333MHz
-MSI 790GX-G65
-CM V8 CPU cooler
-Corsair TX650
-Samsum F3 1TB and Seagate Barracuda 500GB
-CM HAF 922
fan configuration - intake: 200-front, 120-botom, 80-window. output: 200-top, 120-back
Both cards use the same Afterburner fan speed vs temp graph:
Ambient temperature sensor: Fluke 51 series, with probe positioed betweem VGA and CPU cooler
Initial ambients are about 21-23C. I opened my window after every test to get the hot air out before doing another test.
Overall setup (note the companion cube plushie)
-Interesting MSI 5850 Twin Frozr specs
All the specs are the same as any reference clock. Though I did find the core voltage to be 1.100V default as compared to my reference card at 1.087V. The fan comes preset to 30% as opposed to reference at 20%.
MSI says it has dual ultra quite PWM fans that provide 50% better thermal efficiency and direct airflow to cool the GPU. It has a large copper base coated with nickel that provides "feastest" heat dissipation (typo on the box). 8mm supper heat pipes. Ultra high-density aluminum heatsink has large area for heat dissipation. Claims 12C cooler under full load (shows a graph of the frozr vs a reference with the frozr at 76C and the reference at 88C, states "Results is based on 25C of system temperature). It has military grade componets including, highly conductive caps, solid caps, and solid state chokes.
-Cards to be compared
Sapphire 5850 reference card and MSI 5850 Twin Frozr
-Tests
Both cards are being tested with stablity and benchmarks at stock settings and overclocked to 840MHz on the core and 1150MHz on the memory at 1.112V on the core. I recorded the ambient temperature using my Fluke, and I used GPU-Z 0.4.0 and its sensor logging to record the GPU temps during th test.
Tests and settings
-FurMark 1.7.0, run for 5 minutes
Run Mode: Stability Test, checked Xtreme Burning Mode and Log GPU temperature
Window Params: Fullscreen, 1920x1080, MSAA 8X
-Unigine Heaven 2.0
API: dx11
Shaders: high
Tessellation: extreme
Anisotropy: 16
Anti-aliasing: 8x
Fullscreen, 1920x1080
-Call of Pripyat
Preset: ultra
Renderer: dx11
1920x1080
MSAA: 4x
MSAA for A-tested objects: dx10.1 style
SSAO: HDAO
SSAO Quality: ultra
checked: use dx10.1, enable tessellation, contact hardening shadows
-Far Cry 2 benchmark tool
map: small ranch
1920x1080, fullscreen, vsync
Performance: all very high
Overall Quality: ultra high
-GPUTool 0.0.0, run for 5 minutes
checked: sync clocks of multi-GPU, renderer uses fullscreen
everything else default
-Dirt2 benchmark
1920x1080, 60Hz, vsync
MSAA: 8x
everything else turned to highest setting
-FPS Results
-Heaven
Sapphire stock
Sapphire OCed
MSI stock
MSI OCed
-Call of Pripyat
Sapphire stock
Sapphire OCed
MSI stock
MSI OCed
-Far Cry 2
Sapphire stock
Sapphire OCed
MSI stock
MSI OCed
-Dirt 2
Data pulled form the XML dirt 2 produces
-Temperature Results
Only the OCed temps are in the graphs, mainly because I think they are the more important ones.
-Notes on tests
EDIT: Each card has now done each test on the same drivers, Cat 10.3 and Afterburner 1.5.1
-Result Discussion
I initially bought the MSI card due to its advertised coolness, aka 12% less the reference, and quiter fan speed. The fan did turn out to be much quiter. Up to about 60-70% fan speed, the sound level is less than that of the reference. The temperature results of the MSI are a bit unsettling to me. While I see that the graph posted on the box is for very high temps (70C+), I would think I would see even slightly better temps at lower loads. Things get even worst with OCing.
Clock for clock comparison to the Sapphire, there is no noticeable FPS gain that I can tell. I expected this, and thus really tried to focus on the temps camparision.
The Newegg advertisement says "OverVoltage Support by using latest AfterBurner 1.5.1". I have Afterburner set up correctly so I can control the voltages and such. At 1.112V, the card was fine. At 1.150V the card freaks out under load, regardless of the core and mem clock frequencies. Starting GPUTool causes the screen to go black, and eventually stops giving a signal, and the fan speed goes over 60% or so. I got through 25 stages of Heaven with 1.149V, but artifacted like crazy on the last stage. So far I am not happy.
I am in contact with MSI about these issues. Hopefully they get resolved. I may eventually flash the BIOS and try going from there. But for now, this card is not worth the $329 I spent. MSI has since emailed me back offering no help since "we do not support overclocking because it puts ware and tear on the parts. Also there is no guarantees that overclocking it will work".
-MSI With cardboard shroud test results
Added a shroud to see if I could reduce the temps by more efficiently removing the hot air out the back. Without the shroud, the air from the fans will just enter the case. The shroud is made from a cereal box and cover the whole card with holes for the fans and held on with scotch tape. Also used the tape to seal other openings, though some still exist.
Ran the same battery of tests again in the same testing conditions. Temp results:
Very slight improvements in GPU temps, but big improvement in ambient temps. This is most likely caused by less leakage into the case.
-Final thoughts
Over the course of the test, the fan noise was very minimal compared to the reference card. I have a few conclusions on this that also relate to the temperature:
1. The fan blades are significantly smaller than other cooling fans on other cards that are positioned on top of the heat sinks. Thus, there is less airflow and less noise produced by the fans.
2. Assuming #1 is right, then less airflow would mean worst temps compared to the reference. It seems to me that my tests support this.
Some how, MSI was able to produce temps of 76C on the Frozr and 88C on a reference, thus allowing them to say a 12C cooler card. This seems a little extreme to me, since none of my tests gave temps of this magnitude, though I am using a different setup and conditions (MSI doesn't state theirs).
If you are looking for a cooler card, I would say stay away and get a water block. If you want a quieter card, then this should be a consideration.
