Hey everyone,
I thought I'd start up an ongoing thread to chronicle my progress with my AMD A10-5800K APU project.
Of important note: I built this machine as a budget computer, to connect to my TV to play games from my couch at 720p resolution at medium to high graphics settings (depending on the framerate I can get out of any given game based on its individual graphics settings).
Starting out as a complete tower build (no keyboard/mouse/monitor) for just under $350:
Processor: AMD A10-5800K APU @ 3.8GHz (stock)
Motherboard: MSI FM2-A75MA-E35
RAM: G.Skill Ares 8GB (2x 4GB) DDR3 @ 1866MHz
Graphics: APU integrated Radeon HD 7660D (1GB allocated system memory)
PSU: Cooler Master eXtreme Power Plus RS-550-PCAR-E3 550w
HDD: Seagate Barracuda 1TB 7200rpm SATA 6.0Gb/s
Case: Cooler Master Elite 311 ATX Mid Tower
OS: Microsoft Windows 8 Pro x64 (full version 'upgrade' - can clean install with no issues)
(yes, the computer lacks an optical drive, but being used primarily as a gaming machine, I found no need to spend the money for one even though they are dirt cheap)
Since the initial build of this machine, I have reverted back to Windows 7 Home Premium x64 which I had previously owned (just was not a fan of Windows 8) and have added a Cooler Master Seidon 120M Closed Loop liquid cooling system (washing each other out at $40 a piece, so technically I'm still under $350 on this build even with the liquid cooler).
A few things to note, specifically with the Cooler Master Elite 311 case:
1. The radiator did not sit flush against the case itself to allow for a push fan configuration due to a rivet ever so slightly blocking the bottom corner of the radiator from sitting flush against the back of the case (however based on many benchmarks done with push vs. pull cooling over heat fins, and with a pull configuration coming in even or sometimes 1-2 degrees lower than a push configuration, I was happy to set mine up in a pull configuration with the fan between the case and the radiator).
2. The Case panel has metal tubing around 3 edges of the panel, which stick out farther than the securing slats along the top and bottom of the side panel. What this means is that if you have an component (in my case, the radiator) sitting very close to the edge of the case in a location where it would interfere with the metal tubing, the side panel will not go on. To remedy this (after attempting to just squeeze the tubing flat with pliers to no avail), I ended up using bolt cutters (the metal tubing is extremely solid) to snip away at the part of the tubing that was up against the radiator, cutting as much as I could away, and then using pliers to flatten out the remaining pieces that I coudn't snip away (I then covered with electrical tape to prevent any cuts/scrapes from the jagged metal).
3. The peg for the case window closest to the radiator was also in the way, so I cut off the bottom of the peg (so that it no longer stuck through the plastic window) and simply taped that portion of the window to the side panel (as you can see in the picture above), and then superglued the remaining portion of the peg to the front side of the panel for aesthetics.
Here is a picture with the Cooler Master Seidon 120M installed (with the fan set up in a pull configuration venting the air out of the back of the case) along with the side panel modification I performed (I definitely could have achieved a much cleaner cut with a power tool, but I opted not to):
The stock fan, at full bore (~2400 rpm) is quite loud (Rated ~40 dBA), but it does a great job at cooling, keeping my system ~27 degrees C idle, compared to ~40 degrees C idle with the stock AMD heatsink & fan (granted the radiator exhaust fan @ 2400 rpm is louder than the stock AMD fan). Due to the noise level of the included Cooler Master 120MM Radiator fan at full blast, I decided to try out the Cougar Vortex CF-V12HPB Hydro-Dynamic-Bearing (Fluid) 12CM Silent Cooling Fan with Pulse Width Modulation (this bumps the build up to $355, however it is feasible that I could sell the stock Cooler Master fan for ~$5 bringing the total cost back down to a cool $350). The Cougar boasts 800-1500 rpm with 70.5 CFM @ 17.9 dBA, whereas the Cooler Master weighs in at 600-2400 rpm with 86.15 CFM @ a whopping 40 dBA.
Installed the Cougar CF-V12HPB fan in pull configuration mounted between the case and the radiator. The fan is rated at 800-1500rpm @ 17.3dB. My motherboard, when set to run the fan at full load, gets 1700rpm out of it. Ends up not much different (~2 degrees C) between the stock Cooler Master fan and the new Cougar fan at idle temps at each fan's full speed. Here's a graph at full CPU load showing fan percentage settings and correlating temperature (top lines) in Degrees Celcius after running Prime95 Large FFT for 30 minutes (for each datapoint), along with a delta in decibel increase (bottom lines) at each fan percentage setting:
A quick note about the fan speeds in the graph above, the MSI motherboard does not allow for a specified rpm but it allows for 'minimum speed percentages' (which will then automatically increase in speed when the 'target' CPU temp is reached). For this test, I set the 'target' CPU temp to 70 degrees C (with the notion that I would not reach that temp and therefore the fan would not vary in speed for each test). The fan speeds (RPMs) are not listed because they do not line up with each other (since the max fan speeds are different).
Based on the fan test results, I have re-installed the stock Cooler Master fan (bumping the build back down to under $350), as it ends up actually doing a great job for cooling at very manageable noise levels. Its sweet spot seems to lay ~50% fan speed with regards to cooling efficiency and noise level. Also, it's nice to know that I have extra overhead for cooling when I get set to overclock.
Now, on to overclocking! First I tried MSI's 'OC Genie' mode. My hardware does not like this mode! 3DMark11 and Batman Arkham Asylum both crash (I'm not going any further with the 'OC Genie' mode, due to this). I believe the issue is that 'OC Genie' does not increase voltages at all, and when it bumps up FSB and GPU Clock speeds it runs into issues.
I gave it a go manually and came up with the following that got me 'stable' with the best 3DMark11 scores:
CPU Base Frequency (MHz) 100
CPU Ratio 45
GPU Engine Frequency 950MHz
CPU Voltage 1.550000v
CPU-NB Voltage 1.349400v
DRAM Frequency 2133MHz
DRAM Voltage 1.650v
And with that, I'm 'stable' at 4.5GHz:
For the first of the gaming benchmarks, I ran 3DMark11 (free version so my configuration options were zero), scored the following on the 'Performance' setting (1280 x 720) both stock and at my current max overclock:
I feel like this shows a pretty solid gain! I'm happy with the results so far.
Interesting note, I had originally allocated 2GB system RAM to the GPU, and I was curious how reducing down to 1GB allocated would affect the benchmarks. To my surprise, in 3DMark11, the scores were better (going from 1577 @ 2GB to 1584 @ 1GB). I'm not quite sure why this is (if anything, I would have expected the score to stay the same or drop), but possibly the extra system memory assisted in boosting the score whereas it went unused when allocated to the GPU. Based on this, I ran benchmarks for Batman Arkham Asylum and Batman Arkham City at both 2GB and 1GB allocated, and the results were basically spot on, ~1-2FPS +/- no matter if it was 2GB or 1GB allocated. Therefore, I have dropped my GPU RAM allocation to 1GB (all benchmarks in this thread represent values at 1GB allocation).
And now on to individual game benchmarks (please note that the Y-Axis scaling in the following graphs are not standardized. Read the numbers rather than relying only on the visual size representation).
Batman Arkham Asylum:
Batman Arkham City:
Crysis:
As of right now, the following benchmarks are all at stock hardware speeds. They will be updated as I continue to benchmark under overclocked conditions
Lost Planet 2 (Test B):
I thought I'd start up an ongoing thread to chronicle my progress with my AMD A10-5800K APU project.
Of important note: I built this machine as a budget computer, to connect to my TV to play games from my couch at 720p resolution at medium to high graphics settings (depending on the framerate I can get out of any given game based on its individual graphics settings).
Starting out as a complete tower build (no keyboard/mouse/monitor) for just under $350:
Processor: AMD A10-5800K APU @ 3.8GHz (stock)
Motherboard: MSI FM2-A75MA-E35
RAM: G.Skill Ares 8GB (2x 4GB) DDR3 @ 1866MHz
Graphics: APU integrated Radeon HD 7660D (1GB allocated system memory)
PSU: Cooler Master eXtreme Power Plus RS-550-PCAR-E3 550w
HDD: Seagate Barracuda 1TB 7200rpm SATA 6.0Gb/s
Case: Cooler Master Elite 311 ATX Mid Tower
OS: Microsoft Windows 8 Pro x64 (full version 'upgrade' - can clean install with no issues)
(yes, the computer lacks an optical drive, but being used primarily as a gaming machine, I found no need to spend the money for one even though they are dirt cheap)
Since the initial build of this machine, I have reverted back to Windows 7 Home Premium x64 which I had previously owned (just was not a fan of Windows 8) and have added a Cooler Master Seidon 120M Closed Loop liquid cooling system (washing each other out at $40 a piece, so technically I'm still under $350 on this build even with the liquid cooler).
A few things to note, specifically with the Cooler Master Elite 311 case:
1. The radiator did not sit flush against the case itself to allow for a push fan configuration due to a rivet ever so slightly blocking the bottom corner of the radiator from sitting flush against the back of the case (however based on many benchmarks done with push vs. pull cooling over heat fins, and with a pull configuration coming in even or sometimes 1-2 degrees lower than a push configuration, I was happy to set mine up in a pull configuration with the fan between the case and the radiator).
2. The Case panel has metal tubing around 3 edges of the panel, which stick out farther than the securing slats along the top and bottom of the side panel. What this means is that if you have an component (in my case, the radiator) sitting very close to the edge of the case in a location where it would interfere with the metal tubing, the side panel will not go on. To remedy this (after attempting to just squeeze the tubing flat with pliers to no avail), I ended up using bolt cutters (the metal tubing is extremely solid) to snip away at the part of the tubing that was up against the radiator, cutting as much as I could away, and then using pliers to flatten out the remaining pieces that I coudn't snip away (I then covered with electrical tape to prevent any cuts/scrapes from the jagged metal).
3. The peg for the case window closest to the radiator was also in the way, so I cut off the bottom of the peg (so that it no longer stuck through the plastic window) and simply taped that portion of the window to the side panel (as you can see in the picture above), and then superglued the remaining portion of the peg to the front side of the panel for aesthetics.
Here is a picture with the Cooler Master Seidon 120M installed (with the fan set up in a pull configuration venting the air out of the back of the case) along with the side panel modification I performed (I definitely could have achieved a much cleaner cut with a power tool, but I opted not to):
The stock fan, at full bore (~2400 rpm) is quite loud (Rated ~40 dBA), but it does a great job at cooling, keeping my system ~27 degrees C idle, compared to ~40 degrees C idle with the stock AMD heatsink & fan (granted the radiator exhaust fan @ 2400 rpm is louder than the stock AMD fan). Due to the noise level of the included Cooler Master 120MM Radiator fan at full blast, I decided to try out the Cougar Vortex CF-V12HPB Hydro-Dynamic-Bearing (Fluid) 12CM Silent Cooling Fan with Pulse Width Modulation (this bumps the build up to $355, however it is feasible that I could sell the stock Cooler Master fan for ~$5 bringing the total cost back down to a cool $350). The Cougar boasts 800-1500 rpm with 70.5 CFM @ 17.9 dBA, whereas the Cooler Master weighs in at 600-2400 rpm with 86.15 CFM @ a whopping 40 dBA.
Installed the Cougar CF-V12HPB fan in pull configuration mounted between the case and the radiator. The fan is rated at 800-1500rpm @ 17.3dB. My motherboard, when set to run the fan at full load, gets 1700rpm out of it. Ends up not much different (~2 degrees C) between the stock Cooler Master fan and the new Cougar fan at idle temps at each fan's full speed. Here's a graph at full CPU load showing fan percentage settings and correlating temperature (top lines) in Degrees Celcius after running Prime95 Large FFT for 30 minutes (for each datapoint), along with a delta in decibel increase (bottom lines) at each fan percentage setting:
A quick note about the fan speeds in the graph above, the MSI motherboard does not allow for a specified rpm but it allows for 'minimum speed percentages' (which will then automatically increase in speed when the 'target' CPU temp is reached). For this test, I set the 'target' CPU temp to 70 degrees C (with the notion that I would not reach that temp and therefore the fan would not vary in speed for each test). The fan speeds (RPMs) are not listed because they do not line up with each other (since the max fan speeds are different).
Based on the fan test results, I have re-installed the stock Cooler Master fan (bumping the build back down to under $350), as it ends up actually doing a great job for cooling at very manageable noise levels. Its sweet spot seems to lay ~50% fan speed with regards to cooling efficiency and noise level. Also, it's nice to know that I have extra overhead for cooling when I get set to overclock.
Now, on to overclocking! First I tried MSI's 'OC Genie' mode. My hardware does not like this mode! 3DMark11 and Batman Arkham Asylum both crash (I'm not going any further with the 'OC Genie' mode, due to this). I believe the issue is that 'OC Genie' does not increase voltages at all, and when it bumps up FSB and GPU Clock speeds it runs into issues.
I gave it a go manually and came up with the following that got me 'stable' with the best 3DMark11 scores:
CPU Base Frequency (MHz) 100
CPU Ratio 45
GPU Engine Frequency 950MHz
CPU Voltage 1.550000v
CPU-NB Voltage 1.349400v
DRAM Frequency 2133MHz
DRAM Voltage 1.650v
And with that, I'm 'stable' at 4.5GHz:
For the first of the gaming benchmarks, I ran 3DMark11 (free version so my configuration options were zero), scored the following on the 'Performance' setting (1280 x 720) both stock and at my current max overclock:
I feel like this shows a pretty solid gain! I'm happy with the results so far.
Interesting note, I had originally allocated 2GB system RAM to the GPU, and I was curious how reducing down to 1GB allocated would affect the benchmarks. To my surprise, in 3DMark11, the scores were better (going from 1577 @ 2GB to 1584 @ 1GB). I'm not quite sure why this is (if anything, I would have expected the score to stay the same or drop), but possibly the extra system memory assisted in boosting the score whereas it went unused when allocated to the GPU. Based on this, I ran benchmarks for Batman Arkham Asylum and Batman Arkham City at both 2GB and 1GB allocated, and the results were basically spot on, ~1-2FPS +/- no matter if it was 2GB or 1GB allocated. Therefore, I have dropped my GPU RAM allocation to 1GB (all benchmarks in this thread represent values at 1GB allocation).
And now on to individual game benchmarks (please note that the Y-Axis scaling in the following graphs are not standardized. Read the numbers rather than relying only on the visual size representation).
Batman Arkham Asylum:
Batman Arkham City:
Crysis:
As of right now, the following benchmarks are all at stock hardware speeds. They will be updated as I continue to benchmark under overclocked conditions
Lost Planet 2 (Test B):
Last edited: