Overclock.net banner

1 - 20 of 83 Posts

·
Premium Member
Joined
·
4,727 Posts
Discussion Starter #1
SAPPHIRE Radeon HD 5850 Review



Table of Contents:
1. Introduction, Unboxing, Specification, Features, Disassembly
2. Drivers, Temperatures, Overclocking, Testbed
3. Benchmarks 1
4. Benchmarks 2
5. Benchmarks 3
6. Benchmarks 4
7. Direct Compute, GPGPU, PCI-E Scaling, Conculsion

Introduction:

Today we take a look at the Sapphire Radeon HD 5850 1GB video card from
ATi; the second DX11 compatable card from the newly released Radeon HD
5000 series. The 5850 uses the RV870 core, and like all gpus in the line, is
manufactured using 40nm process.

It is close in comparison to its bigger brother, the HD 5870. Sporting the
same 2.15 billion transistors count, 334 mm2 die size, 32 ROP's, and 1GB of
Hynix 1250mhz rated GDDR5 memory. The differernces between the 5870
and the 5850 are the shader count and the clock speeds.

Radeon HD 5850 has 9 x 160 Shader clusters = 1440 Shader processors
Radeon HD 5870 has 10 x 160 Shader clusters = 1600 Shader processors

Radeon HD 5850 has a 725 mhz core speed and memory clocked at 1000 mhz.
Radeon HD 5870 has a 850 mhz core speed and memory clocked at 1200 mhz.

Up to this point with the HD 4000 and now HD 5000 series, ATi has done a
great job of covering the whole price and performance range of graphic
cards. Therefore spec-wise, as the HD 5870 is the successor and
comparable to the 4870 X2, so the HD 5850 should be closely comparable
to the 4850 X2. So let's find out what the Sapphire HD 5850 will do.

Unboxing:













Specification:
Process 40nm
Transistors 2.15 Billion
Engine Clock 725MHz
Stream Processors 1440
Compute performance 2.09TFlops
Texture Units 72
Texture Fillrate 53.2 Gtexel/s
ROP 32
Pixel Fillrate 23.2 Gpixel/s
Z/Stencil 92.8Gsamples/s
Memory Type GDDR5
Memory Clock 1000Mhz
Memory Data Rate 4.0 GB/s
Memory bandwidth 128.0 GB/s
Maximum Board Power 170watts
Idle Board Power 27watts





Features:
*Microsoft DirectX® 11 Support
*ATI Eyefinity Technology
*ATI Stream Technology
*Designed for DirectCompute 5.0 and OpenCL
*Accelerate Video Transcoding
*Advanced GDDR5 Memory Technology
*2nd Generation TeraScale Engine
*Microsoft Windows 7® Support
*ATI CrossFireX™ Technology
*Enhanced Anisotropic Filtering
*Accelerated Video Transcoding
*Display Flexibility, Supports DL-DVI, DP, HDMI and D-Sub
*HDMI 1.3
*Dolby® TrueHD and DTS-HD Master Audio™ Support
*ATI PowerPlay™ Technology - Enhanced Support for GDDR5 memory
*ATI Avivo™ Technology Enhanced Unified Video Decoder 2 (UVD 2)
*Supports OpenGL 3.1

Disassembly:











A very well made card, solid contruction, killer cooler and as you
can see a heatpipe across the VDDC.
 

·
Premium Member
Joined
·
4,727 Posts
Discussion Starter #2
Drivers:

Catalyst Control Center - Graphics Software Info


Catalyst Control Center - Hardware Info


Catalyst Control Center - Overdrive


Temperatures:

Idle temperture


Load temperature - Fan on auto


Load temperature - Fan at 50%


Load temperature - Fan at 100%


Overclocking:

Overclocking the HD 5850 is fairly simple and can be done a few ways:

-Use Catalyst Control Center Overdrive, which is capped at 775 mhz core/1175 mhz mem.
Since this method is very limited, most users use one of the other two methods.

-Flash the bios to an unlocked or higher clock limited bios, such as the MSI and
ASUS HD5850 bios. If flashing is not your cup of tea or you are like me, where you
only want high overclocks for benching, you can use the last method.

-Use AMD GPU Tool for 5850 to increase both core and memory clocks passed CCC
limits and use MSI Afterburner to increase the voltage.

I chose the last method and was about to reach a stable 1Ghz core clock and 1250 mhz
on the memory clock. One area I would like a future tool to address is a voltage control for
the memory. I tapped out at 1035 mhz core by lowering the memory clock to 1220 mhz.
If I was able to increase the memory voltage, I believe the card would gain for stability and
would be able to be pushed to 1100 mhz on the core. Nevertheless, I fine the incredible
overclocking headroom of the HD 5850 to be amazing.

Tools used:
-AMD GPU Clock Tool for HD 5850/5780
-MSI Afterburner

HD 5850 at 1000 mhz gpu / 1250 mhz mem


Testbed:
E8300 @ 3.8G
ASUS P5Q-PRO
6GB DDR2
ATI Catalyst 9.11
SAPPHIRE Radeon 5850 1GB (stock)(725 mhz core / 1000 mhz mem)
SAPPHIRE Radeon 5850 1GB (Overclocked at 850 mhz gpu / 1200 mhz mem) (5870 clock speeds)
SAPPHIRE Radeon 5850 1GB (Overclocked at 1000 mhz gpu / 1250 mhz mem)
SAPPHIRE Radeon 4850 X2 2GB (Stock)(625 mhz core / 993 mhz mem)





Testing:
3DMark 06
3DMark Vantage
Call of Duty: World at War
Call of Juarez
Cinebench
Crysis
Devil May Cry 4
Dirt 2
Enemy Territory: QUAKE Wars
FarCry 2
Furmark
Left 4 Dead
Resident Evil 5
Race Driver: GRID
Shattered Horizons
Street Fight IV
S.T.A.L.K.E.R.: Call of Pripyat
S.T.A.L.K.E.R Clear Sky
The Last Remnant
Tom Clancy: H.A.W.X.
Unigine Heaven
Unigine Sanctuary
X3 Terran Conflict
 

·
Premium Member
Joined
·
4,727 Posts
Discussion Starter #3
Benchmarks:



3DMark 06
5850 = 18313
5850 850mhz = 18729 (2% gain over stock)
5850 1Ghz = 19092 (4% gain over stock)
4850 X2 = Not tested









3Dmark Vantage: Performance
5850 = P11253
5850 850mhz = P12225 (8% gain over stock)
5850 1Ghz = P13229 (17% gain over stock)
4850 X2 = P11054









3Dmark Vantage: High
5850 = H8730
5850 850mhz = H9905 (13% gain over stock)
5850 1Ghz = H10923 (25% gain over stock)
4850 X2 = H8172











Call of Duty: World at War
1680 x 1050, 4xAA, 16xAF, DX9, Maximum Detail, Recorded demo
5850 = 127 fps average
5850 850mhz = 145 fps average (14% gain over stock)
5850 1Ghz = 156 fps average (22% gain over stock)
4850 X2 = 137 fps average

1920 x 1080, 4xAA, 16xAF, DX9, Maximum Detail, Recorded demo
5850 = 117 fps average
5850 850mhz = 132 fps average (12% gain over stock)
5850 1Ghz = 148 fps average (26% gain over stock)
4850 X2 = 128 fps average











Call of Juarez:
1680 x 1050, 4xAA, DX10, Highest Quality Settings
5850 = 63 fps average
5850 850mhz = 73 fps average (15% gain over stock)
5850 1Ghz = 79 fps average (25% gain over stock)
4850 X2 = 62 fps average









1920 x 1080, 4xAA, DX10, Highest Quality Settings
5850 = 55 fps average
5850 850mhz = 66 fps average (20% gain over stock)
5850 1Ghz = 74 fps average (34% gain over stock)
4850 X2 = 55 fps average











Cinebench:
OpenGL
5850 = 9040
5850 850mhz = 9048 (0.08% gain over stock)
5850 1Ghz = 9106 (0.7% gain over stock)
4850 X2 = 9003











Crysis:
1600 x 1200, DX10, 64 bit, Very High, 0xAA
5850 = 34 fps average
5850 850mhz = 37 fps average (8% gain over stock)
5850 1Ghz = 44 fps average (29% gain over stock)
4850 X2 = 35 fps average

Crysis:
1900 x 1200, DX10, 64 bit, Very High, 0xAA
5850 = 30 fps average
5850 850mhz = 34 fps average (13% gain over stock)
5850 1Ghz = 37 fps average (23% gain over stock)
4850 X2 = 30 fps average

Crysis:
1920 x 1080, DX10, 64 bit, Very High, 4xAA
5850 = 27 fps average
5850 850mhz = 31 fps average (14% gain over stock)
5850 1Ghz = 34 fps average (25% gain over stock)
4850 X2 = 27 fps average







 

·
Premium Member
Joined
·
4,727 Posts
Discussion Starter #4
Benchmarks:



Devil May Cry 4:
1600 x 1200, DX10, 8xAA, 16xAF, Highest Quality Settings
5850 = 141 fps average
5850 1Ghz = 176 fps average (24% gain over stock)
4850 X2 = 157 fps average









Dirt 2:
1920 x 1080, DX11, 8xAA, Ultra Quality Settings
5850 = 47 fps average
5850 1Ghz = 58 fps average (23% gain over stock)
4850 X2 = N/A







Enemy Territory: QUAKE Wars
1680 x 1050, DX10, 16xAF, 4xAA, Highest Quality Settings
5850 = 112 fps average
5850 1Ghz = 114 fps average (1% gain over stock)
4850 X2 = 102 fps average

1920 x 1080, DX10, 16xAF, 4xAA, Highest Quality Settings
5850 = 104 fps average
5850 1Ghz = 110 fps average (5% gain over stock)
4850 X2 = 91 fps average









FARCRY 2:
1680 x 1050, DX10, Ultra High Settings, 16xAF, 4xAA
5850 = 66 fps average
5850 1Ghz = 76 fps average (15% gain over stock)
4850 X2 = 63 fps average







1920 x 1080, DX10, Ultra High Settings, 16xAF, 4xAA
5850 = 61 fps average
5850 1Ghz = 73 fps average (19% gain over stock)
4850 X2 = 57 fps average









Furmark:
1680 x 1050, OpenGL, 8xAA
5850 = 37 fps average
5850 1Ghz = 42 fps average (13% gain over stock)
4850 X2 = 34 fps average



1920 x 1080, OpenGL, 8xAA
5850 = 35 fps average
5850 1Ghz = 39 fps average (11% gain over stock)
4850 X2 = 32 fps average





Left 4 Dead:
1680 x 1050, DX9, 16xAF, 8xAA, Highest Quality Settings, Recorded demo
5850 = 150 fps average
5850 1Ghz = 162 fps average (8% gain over stock)
4850 X2 = 143 fps average

1920 x 1080, DX9, 16xAF, 8xAA, Highest Quality Settings, Recorded demo
5850 = 139 fps average
5850 1Ghz = 153 fps average (10% gain over stock)
4850 X2 = 136 fps average





 

·
Premium Member
Joined
·
4,727 Posts
Discussion Starter #5
Benchmarks:



Resident Evil 5:
1680 x 1050, DX10, 16xAF, 8xAA, Highest Quality Settings, Motion Blur Enabled
5850 = 83 fps average
5850 1Ghz = 100 fps average (20% gain over stock)
4850 X2 = 87 fps average







1920 x 1080, DX10, 16xAF, 8xAA, Highest Quality Settings, Motion Blur Enabled
5850 = 75 fps average
5850 1Ghz = 92 fps average (22% gain over stock)
4850 X2 = 78 fps average









Race Driver GRID:
1920 x 1080, DX9, 16xAF, 8xAA, Highest Quality Settings, (3) Timed Races
5850 = 111 fps average
5850 1Ghz = 125 fps average (12% gain over stock)
4850 X2 = 115 fps average









Shattered Horizon:
1920 x 1080, DX10, 16xAF, 8xAA, Highest Quality Settings, (4) Timed Sessions
5850 = 36 fps average
5850 1Ghz = 50 fps average (38% gain over stock)
4850 X2 = 38 fps average









Street Fighter IV:
1920 x 1080, DX9, 16xAF 8xAA, Highest Quality Settings
5850 = 78 fps average
5850 1Ghz = 99 fps average (26% gain over stock)
4850 X2 = 118 fps average









S.T.A.L.K.E.R. Call Of Pripyat:
1920 x 1080, DX10.1, 16xAF, 4xAA, Preset: Ultra
5850 = 40 fps average
5850 1Ghz = 49 fps average (22% gain over stock)
4850 X2 = 36 fps average







1920 x 1080, DX11, 16xAF, 4xAA, Preset: Ultra
5850 = 41 fps average
5850 1Ghz = 52 fps average (26% gain over stock)
4850 X2 = N/A







S.T.A.L.K.E.R. Clear Sky:
1680 x 1050, DX10.1, 16xAF, 4xAA, Preset: Ultra
5850 = 35 fps average
5850 1Ghz = 45 fps average (28% gain over stock)
4850 X2 = 35 fps average







1920 x 1080, DX10.1, 16xAF, 4xAA, Preset: Ultra
5850 = 31 fps average
5850 1Ghz = 41 fps average (32% gain over stock)
4850 X2 = 30 fps average





 

·
Premium Member
Joined
·
4,727 Posts
Discussion Starter #6
Benchmarks:



The Last Remnant:
1920 x 1080, DX9, 0xAA
5850 = 103 fps average
5850 1Ghz = 123 fps average (19% gain over stock)
4850 X2 = 99 fps average









Tom Clancy: H.A.W.X.:
1680 x 1050, DX10.1, 16xAF, 8xAA, Highest Quality Settings
5850 = 66 fps average
5850 1Ghz = 78 fps average (18% gain over stock)
4850 X2 = 75 fps average







1920 x 1080, DX10.1, 16xAF, 8xAA, Highest Quality Settings
5850 = 59 fps average
5850 1Ghz = 72 fps average (22% gain over stock)
4850 X2 = 60 fps average









Unigine Heaven:
1680 x 1050, DX10, 16xAF, 8xAA, Highest Quality Settings
5850 = 33 fps average
5850 1Ghz = 39 fps average (18% gain over stock)
4850 X2 = 32 fps average







1920 x 1080, DX10, 16xAF, 8xAA, Highest Quality Settings
5850 = 29 fps average
5850 1Ghz = 36 fps average (24% gain over stock)
4850 X2 = 28 fps average







1920 x 1080, DX11, 16xAF, 8xAA, Highest Quality Settings, Tessellation on
5850 = 17 fps average
5850 1Ghz = 21 fps average (23% gain over stock)
4850 X2 = N/A







Unigine Sanctuary:
1920 x 1080, DX10, 16xAF, 8xAA, Highest Quality Settings
5850 = 94 fps average
5850 1Ghz = 121 fps average (28% gain over stock)
4850 X2 = 92 fps average









X3 Terran Conflict:
1920 x 1080, DX10, 16xAF, 8xAA, Highest Quality Settings
5850 = 89 fps average
5850 1Ghz = 91 fps average (2% gain over stock)
4850 X2 = 91 fps average





 

·
Premium Member
Joined
·
4,727 Posts
Discussion Starter #7
Direct Compute and GPGPU:

Direct Compute is application programming interface developed my Microsoft, incorporating "parallel
computing". Similar to Open CL, Direct Compute uses the processing power of a gpu to run many pc
calculation simultaneously. Performing like your typical multi-core cpu, Direct Compute, Open CL
and other GPGPU framework interfaces can take advantage and utilize the 1400 shader processors
the HD 5850 has. How does this apply to gaming? Well all of these application programming interfaces
can and are used to accelerate game physics, AI and improve video transcoding.

Let's test: HD 5850 gpu (stock) vs. Intel E8300 dual core cpu (3.8G)

Single Thread



Multi Thread



SiSoftware Sandra 2010 GPGPU Benchmark

GPGPU Processing - STREAM Processor


GPGPU Processing - Compute Shader Processor


GPGPU Memory Bandwidth - STREAM Processor


GPGPU Memory Bandwidth - Compute Shader Processor


PCI-E Scaling: 16X vs. 8X

Saturating (some like to use the term "bottlenecking") the pci-e port has always been a
concern for those us still using P45 or P35 and are either running cards with pci-e 1.0 or pci-e 2.0
8X/8X crossfire motherboards. Matter of fact, those subject is one of the most heated ongoing debate on
the internet concerning motherboards and video cards. On another thread, fellow OCN members
Almighty, Phil_Boggild, and myself, tested and confirmed most of the 4800 series on this very issue. So, I
though it would be a good idea to include similar comparison test for the HD 5850.

16X


8X (2% or less loss compared to 16X)


16X


8X (2% loss compared to 16X)


16X


8X (1% loss compared to 16X)


Conclusion:

Well I have to say, I'm more than happy with the quality and performance of the Sapphire HD5850 and
it has turned out to be an outstanding card. The performance and overclocking headroom are almost
hard to believe. The temperatures are very good, compared to the 4000 series, the 5850 is an ice
cube. The noise level is relatively low, especially since there is not a whole lot of need to run the fan
over 50%, unless you are benching with an extreme overclock. I give the 5850 a 9 out of 10 just because
the driver performance could still use some work. Hopefully Catalyst 9.12 or Catalyst 10 will increase
the performance of those games titles that need it.

Well that's all she wrote, hope you all enjoyed the review.
Till next card.
 

·
Premium Member
Joined
·
4,727 Posts
Discussion Starter #10
sections 2 and 3 are updated. brb for more, going for coffee.
 

·
Premium Member
Joined
·
4,727 Posts
Discussion Starter #16
Quote:


Originally Posted by tofunater
View Post

Awesome mate. Can't wait to see some results. I'm kinda curious what a tim swap will do for temps on these cards. Have you considered pulling her a part for some nice nekid pics as well as to let us see how much of a mess they made with the stock tim?


I'll see what I can do about disassembling the card and taking some pics.

Update: GPGPU testing added to section 7.
 

·
Registered
Joined
·
3,006 Posts
Awesome review as always! And as I have always said congrats for being ATI's biggest fan! Best review of 5850 yet, + rep. But you really need to do some cable management.
 

·
Premium Member
Joined
·
4,727 Posts
Discussion Starter #20
Quote:


Originally Posted by grunion
View Post

Nice Rico

What's the highest you can get your CD2 benchmark stable?
I'd like to do some direct comparisons since you have a C2D.

cd2? my e8300?
 
1 - 20 of 83 Posts
Top