Original Link: http://www.anandtech.com/show/2283
Gigabyte's New Odin GT 800W Power Supplyby Christoph Katzer on July 24, 2007 12:01 AM EST
- Posted in
Gigabyte is more commonly known for their mainboards and graphics cards. That they also offer power supplies is a new development worth investigating. The Odin series was first announced last year, and people have been waiting since then to see how their offerings perform. Last week we received their 800W model for review and we immediately hooked it up to our torture rack.
There is one feature on this power supply which hasn't been seen before. It is possible to connect this power supply via a USB port to your system and control and monitor it via the included software. Not only can you adjust the fan speeds, but you can also regulate the DC-outputs to a certain extent. In addition you can check on details of the actual voltage distributed and the power drawn from each rail.
The Odin series from Gigabyte is actually built by Channel Well (CWT) which is headquartered in Taoyuan, a suburb of Taipei, Taiwan. They are in the same building as Enermax at the moment, but they are in the process of moving to a bigger facility. CWT has done some OEM production for other big brands in the past as well as power supplies under own brand for system-builders. Last year they started a price-war to get more retail OEM customers and succeeded quite well with it.
Package and Appearance
The power supply comes in a big, colorful box that provides protection during shipping as well as listing the most important features the Gigabyte team has come up with. The package contains two packs of four screws each and all the necessary cables in a bag.
The power supply itself is black and looks anything but ordinary. The back and the side have hexagonal perforations. On the rear this is normal since that's where the air is exhausted. We have not seen these kind of holes in the side, but before anyone thinks that this might interfere with the airflow note that there is a clear plastic sheet right behind the metal. The design on the side is thus for aesthetic reasons only, as far as we can tell.
The power switch is red and contains an LED light. There is no indication on the inlet what voltage range this PSU will handle. This information is written on the package and the side sticker, but it would be useful to include this next to the power input. That way even new users would immediately know what they can connect.
The Odin series has a 13.5cm fan installed which is almost as wide as the power supply itself. With its size it can provide direct airflow to almost any point inside of the housing, though much of the airflow is still blocked by the heatsinks. The larger size also provides for more airflow at lower RPMs, potentially reducing the amount of noise the PSU will generate. We'll look into this more during actual testing.
Cables and Connectors
On the front of the power supply are the jacks for the detachable cables. This is the first modular power supply we've looked at, so we are certainly interested in seeing whether there are any drawbacks to such a design, as has been suggested by other manufacturers in the past. There are four sockets located in the middle that can be used for the Molex/SATA component cables. The jacks on the right are for the two PEG graphics card connectors which come with 6/8 pins. On the left side are jacks for an adjustable fan and four temperature sensors. There are four temperature diodes included, two of which are 50cm and the other two are 70cm in length. Another cable that can be used for a 3-pin fan - for use with the already mentioned port - is also included.
All the cables come in a good quality bag, similar to what we have seen from other high-end manufacturers. The cables are all sleeved, even after the first connector. This is special because it costs quite a bit more to sleeve the cables from end to end, but it results in a more pleasing appearance.
With a minimum length of 50cm and a maximum length of 90cm you can deliver the power to all components even in a large tower chassis. To make comparisons with other power supplies easier we are now using this standard graphic that lists all of the most common lengths. The connectors at the SATA-cables are angled and the last one of the two cables is straight. The ATX connector is very stiff which doesn't allow the cable to move much once connected to the mainboard. If the cable is bent or pulled it could potentially damage the jack on the mainboard.
For the housing of this power supply CWT/Gigabyte came up with a very odd design. Frankly, it took us a few more minutes than usual to open it since it is not done in the normal way. The top, bottom, and right sides (assuming the unit is positioned in a standard ATX case) are a single U-shaped piece that slides away once you remove the screws. With the case open, we are greeted by a layout that is quite different than we normally see.
The three heatsinks are anodized aluminum that only appears to be copper. The heatsinks are flat with lots of fins, and the size of the fins sometimes makes it impossible to see what's underneath. While we did our best, our pictures still can't provide a good look at all of the components located under the heatsinks.
The design of the filtering stage is a typical CWT-layout like we can see in Thermaltake power supplies.
The whole primary side is covered by a heatsink which makes it difficult to identify certain components. Since there is a heatsink on both sides of the primary side, the layout here is slightly different to other power supplies. The big main capacitor sticks out of the heatsink and has a large sticker to make sure everyone can tell where the component is made: Japan.
Some readers might have already recognized the little silver/clear dots on the heatsinks. These are actually LED lights that will shine out of the power supply once it's powered on. Don't worry: the lights can be switched off through the software if you so desire.
There are a total of four LEDs installed in this power supply. They shine through the fan into the system when active. They provide blue lighting and look a little like UV light, but as they aren't true UV lights UV-active material will only glow slightly in the dark.
The secondary side is packed and the components are hard to see. There's a big PCB on the edge of the secondary side that includes all the controls and power connections for the system cables. The fan and temperature diodes also connect to this PCB. There are so many cables that it's impossible for air to come through to cool components under it. Unfortunately, we weren't able to see the capacitors' manufacturer(s) or their rating, but we didn't want to dismantle the unit any further for fear of damaging it prior to testing.
We changed our testing procedures slightly in our last article. We have added a low input voltage test which is in this case again 100VAC. This power supply is rated from 100 to 240VAC thus we couldn't perform a 90VAC test. We are testing the power supplies with programmable loads from Chroma. If you would like to know more about our testing methodology, equipment, and environment, please read our Power Supply Test Methodology overview.
Note: The rails of the power supply can be regulated through the included software. All of our results are made with standard settings without any regulation.
Before we start with the results we would like to clarify the tables from each rail. Some readers were asking after the last review why we wrote 100% load and we had less than the stated 750W. This is easy to explain. In our static tests we are testing a load from 100% and not an output of 100%. We calculate the load for each rail according to the respective combined power and stated power for each line, because the stated power for each line is not the actual power it can deliver together with the other rails in use.
In our review today we have four 12V rails with two of them rated at 18A the other two at 25A. The combined power for these four rails is 62A which is around 15A per rail. The results shown in the tables (in the second column) indicate how much amperage we are using on the rail. The third column shows the resulting wattage, which is the product of the amperage and voltage of that particular line (P = IV). As the voltages drop during testing, the final result is always lower than the actual stated rating of the power supply.
For example: 19.95A is the load for the 3.30V rail at 100%. If the power supply could deliver 19.95A at 3.30V we would have a result of 63.525W and not 60.06W like we will see on the next page. By the time we add this added with the 5V and four 12V rails we get 100% theoretical load at only 770W instead of 800W. This does not mean that the PSU cannot deliver 800W however, and thus we have added more tests towards the end of the article to determine each rail's maximum capacity.
3.3V DC Output
|3.3V DC Output|
(All rails combined)
Starting with the 3.3V rails we encounter slight problems. At lower loads there's nothing to concern us, but as the load rises the 3.3V rail drops tremendously and barely manages to stay within specification. As we will describe later on we have measured the same low voltages with the P-Tuner software which confirms our readings, and we will see this pattern on all of the other rails. It's not only at higher temperatures that this power supply delivers lower voltages; we found even at room temperature where the internal components should stay cooler that we still get large voltage drops.
The heatsinks don't get too hot like as we will see later, but since we only measure the heatsinks we don't know the exact temperatures of the other components on the PCB. As we mentioned earlier, we think that the heatsinks may be too big to let any air through to the components on the board. The dropping voltages could be a result of this.
5V DC Output
|5V DC Output|
(All rails combined)
The 5V rail doesn't look much different from the 3.3V rail. We can see a heavy constant drop for the duration of the test. The results are still within specifications, but it was very close at the highest loads. With a little help from the provided software, however, we could adjust this rail to get it running beautifully close to the ideal output.
5Vsb and Standby Efficiency
|5Vsb DC Output|
(All rails combined)
The standby efficiency of the Odin GT is okay. With lower loads in standby mode we could wish for a better result but it is still acceptable. With no load applied we measured a usage of 0.91W - 1.88W. That's a large amount for doing virtually nothing and GB/CWT should work on improving this.
The voltage of the 5Vsb rail decreases with increasing loads like the other rails, but it still stays within an acceptable range.
12V DC Output
|12V DC Output|
|Percentage||Amperage 12V1||Amperage 12V2||Amperage 12V3||Amperage 12V4||Wattage in total||Wattage
(All rails combined)
With standard settings we were quite worried about the 12V rail since it was already generating lower than expected results even at lower loads. A few of the rails even came close to dropping out of spec. Again, we were able to regulate these rails quite successfully using the P-Tuner software to generate a better result under lower loads. At higher loads the software was less successful at regulating voltages, since we were already using pretty much everything the PSU could put out.
As our latest addition we will now perform dynamic tests on our test power supplies. During normal tests, we use a constant current to bring them to a certain level of load. The load is also constant once we reach that level, which is not necessarily the same as in a normal PC environment. The load generated by a PC is dynamic and can be higher or lower depending on the applications being run. A hard disk for example will suddenly apply a much higher load during startup, as it attempts to bring the platter RPMs up to speed. The more hard disks you have, the higher the startup load will be. These increased loads not only occur at startup but throughout normal usage, and depending on the component generating the load the demands can come on any or all of the rails.
High transient loads (i.e. short spikes of higher power requirements) can be harmful for a power supply and result in the system restarting/rebooting on occasion or sometimes even in damage to the power supply or connected components. The latest Power Supply Design Guide addresses this problem by defining how much transient load should be allowable on any of the rails. Even with high transient loads, the voltages should still stay within the normal specified range.
|Output||Maximum Step Size
(% of rated output amps)
|Maximum Step Size|
With the Chroma test equipment we can apply a very accurate load to the various rails, and in addition we can specify the duration of loading a specified value. This allows us to apply a different load with a different duration for each rail. This is important since there are clear specification differences between each rail. Since the maximum step size for the transient load is rated off the actual performance of each power supply, we have to apply a different transient load level to each new unit that we test.
The transient load will be applied on different stages of constant loading. We felt that testing transient loads at 20%, 50%, and 80% load would be the best way to clearly show meaningful results. The tests will be performed with each of the three different input voltages of 90/100, 115 and 230VAC.
The Gigabyte Odin GT performed very well during our dynamic load tests and was not disturbed by any transient loads. Even an inrush current caused by going from off state to 100% load didn't cause any problems.
In addition to transient testing, we also wanted to look at how the power supplies deal with Over Current Protection (OCP). OCP is an important safety feature that should be present on all power supplies. OCP guards the power supply and the attached components from damage that can occur when too much current is required by any of the rails. In the best case scenario, OCP should kick in and cause the power supply to simply shut down if a rail is stressed to heavily. If this doesn't happen the OPP (Over Power Protection) of our Chroma unit will kick in to save our test equipment. In a real PC, if OCP fails to work the hardware will continue to run until the cables melt, a system crash caused by out of spec voltages changes the demand on the PSU, or in the most extreme cases the power supply explodes. Needless to say, if any of the above happen there is a good chance other hardware could be damaged in the process.
Our OCP test using the Chroma equipment applies a specified amount of load to each rail of the power supply. The test always starts at no load and ends at an amount we set. The highest load always depends on the rated output from the manufacturer - enough that OCP should trigger. If OCP fails to activate, we don't need to worry since we have an OPP installed on the Chroma (and there's no PC equipment attached). We have set the step of each rising load to 2.5A, which will be added every two seconds during the test.
On the graph you will see only the 12V rails. This is because we couldn't apply any load for the OCP test to the lower rails like 3.3V and 5V. When we did so the power supply immediately shut down with no obvious reason. We had to cut the input and couldn't restart the procedure normally immediately after this occurred, so no test on these rails was possible.
As you can see we could apply a load of 24-25A on each rail. This is the usual amount of an OCP, regardless of what the label says about the rated output. There is always a little room for peak load with each power supply. Note that the load from each OCP test is applied to each rail separately. This means that if you wanted to do this test with all rails at the same time it would not be possible without greatly exceeding the rated output of the power supply. The combined power on the label states 744W, which means we could apply around 62A on all the 12V rails at the same time. Divided by four this makes around 15A per rail, which we have seen on a high static load test.
Efficiency and Power Loss
Just last week we reviewed the PC Power & Cooling Silencer 750 QUAD which had a terrific efficiency of up to 86%. This week we see similar results. Even at lower inputs the efficiency reaches 83%, topped by a maximum 85% at 230VAC input. High efficiency is not the only important factor; efficiency at higher loads and over a range of loads is equally (if not more) important. A power supply that reaches 80% efficiency at a 10% load but then drops off at moderate to high loads wouldn't be desirable. Most PSUs are usually loaded at anywhere from 30 to 80% of their rated output, and maintaining a high efficiency at all loads would be the best possible result.
The Odin GT delivers another impressive result, with efficiency that is simply always good. Just after 20% load it is already over 80% efficient and it stays in the 80-85% efficiency range up until 100% load. If your PC has the need for this kind of power supply, you will get better than 80% efficiency and any load above approximately 150W. Considering the 800W rating, we expect this unit to go into systems that idle at 200W or more, so the end result is very good.
Due to the very high efficiency, the power loss is also very low and we see good results.
Power Factor Correction
At lower input voltages the power factor correction again looks very nice. With increasing input like 230VAC we see a tremendous decrease and the correction cannot even reach .99 anymore. This is a normal result, but it could be better for a high-end PSU.
Temperatures (Ambient 25°C)
During room temperature testing we don't see any results that would cause us concern. All three of the heatsinks stay under 40°C. The exhausted air reaches up to 35°C which is a very good result as well.
Temperatures (Ambient 25-50°C)
With our hot ambient temperatures we want to see what the power supplies can really take. We increase the temperature as the graph indicates. At 50°C we have a hotter environment than that used to rate any of the power supplies we test. Most of them will show their first weakness with falling voltages. It will be also more difficult to dissipate all the produced heat from inside. Therefore we monitor the heatsink temperatures, fan speed, and noise to compare all of the results.
All three heatsinks show a good ability to dissipate heat as the exhausted air temperatures are increasing throughout the test. This is very good for the components attached to the heatsinks. As we mentioned before we do feel that the components on the PCB will not get much of the desired airflow and therefore will be hotter; however, we didn't have any temperature diodes attached to other components to prove our theory.
Acoustics and Fan Speed (Ambient 25°C)
Throughout testing we could clearly hear the noise produced by the installed fan, regardless of load, but this is not clearly visible on the decibel graphs. The fan speed begins increasing from around 50% load which equals around 400W of power. Even at a normal ambient temperature the fan began to rotate at a very high speed which resulted in noisy acoustics despite the low temperature. That said, it's worth mentioning that the user can adjust the fan speed to their own needs using the provided software, so if the power supply is being used in a good air-conditioned environment it could surely run at lower speeds and thus lower noise levels. When the PSU is mounted in a system, the noise levels were not as noticeable; this is because the PSU is louder at the fan intake then it is at the exhaust.
Acoustics and Fan Speed (Ambient 25-50°C)
Again, we want to fully stress the power supplies with higher operating environments to see what they can really do. Needless to say, we expect much higher noise levels and fan speeds in these tests.
With the higher ambient temperatures, the fan is almost at its highest speed by the 50% load mark. This is due in part to the target temperature set by the engineers. It can be adjusted via software, but if the PSU reaches 75°C the fan will always be on full speed to try to save the power supply. Again, we could always hear the fan, even at lower loads. At full speed and an operating temperature of 50°C, it's not as loud as the PCP&C Silencer 750, but it's definitely not quiet either.
The P-Tuner Software
We have mentioned the software quite a few times already, so let's take a closer look. The P-Tuner software can be found on the CD. The installation is very easy but it does have one big flaw: it automatically detects the language of the operating system installed on the PC and uses that. You cannot change the language which unfortunately leaves us with the German version of the software since that is the language on our test PC. (Ed: Guten tag?) We might also point out that the translation used for some items is incorrect, and likely that is the same for other languages as well. Hopefully the language issue will be corrected with a patch in the near future, but for now you'll have to deal with German screenshots.
After a successful installation you will be greeted with this screen. This will only occur if the connection between the power supply and system is functioning correctly; if the connection is not established the software fails to start.
On the main-screen we have the actual usage of the system in watts on the left side and the highest amount used during that session on the right side. Compared to our results using the Chroma equipment, the amounts shown aren't identical. For example, with lower load testing we measured 160W from the rails and the software only showed 130W. If that amount had been correct we would have only had 68% efficiency, since the system was drawing 190W from the AC source. At higher loads the software showed 875W and in reality we were only drawing 771W - it's unfortunate that the software was incorrect, as we were about ready to start selling perpetual motion machines. In short, the results of the software are nothing close to reality, but it looks nice and it does at least give home users a place to start.
The second row shows the voltages of each rail. Each rail can be shown by clicking on it. If you want to see all rails at the same time you just click the All ("Alle") button on the lower right side of this row. A second screen will open and show all rails and their respective voltages. We compared the voltage on each rail with our equipment and found the results were quite similar. The differences were in the millivolt range and not more (i.e. less than .01V).
The third row shows the actual power drawn on each rail. By clicking on each rail the load can be seen in the little round display on the right side. The upper amount shows the actual used power on the chosen rail and the lower amount shows the highest power drawn on that rail during the current session. The amounts shown were not been equal to the actual loads we had put on the rails. We measured differences of up to 10A. Bringing up the details display, we can see all rails at once. The display shows on all 12V rails with a maximum 18A which is not quite accurate since two of the rails are rated at 25A.
The P-Tuner Software, Cont'd
The fourth row with the two rotating fans shows the power supply fan on the left and the connected system fan on the right. We have double-checked the RPM with our equipment and were surprised about the high accuracy of this display.
The last row shows the temperature of the power supply itself and the four connected temperature diodes. To make things easier to remember, the user can name each sensor. Note also that the button to switch off the LEDs is located at the lower right side.
A very interesting area is marked with C on the bottom-left of the control-panel. Here you can actually control the power supply and the attached fans. In the top of the newly opened control are two submenus. The first one shows the control for the fans. You can use this to create your own scheme for how the fan should operate in your system. Caution should be exercised when using this utility, as you can force the fan to run very slowly at high temperatures. This can result in a very hot power supply which is certainly not good for any of the components inside and could easily lead to a PSU failure.
In our opinion, the most useful function of all is hidden under this second submenu. You can actually adjust the voltage supplied by the unit. OCZ was first in the market with this idea about four years ago with their PowerStream line of PSUs, but at that time we only got three potentiometers on the rear of the case, and you weren't given any clear indication of the result of any changes. You simply got a green light if the voltage was within spec or an amber light if it was out of spec. With this software you can regulate the output of the rails, and this should definitely be done since the default settings generate poor results as we saw during our tests. You can see the result of your changes immediately back in the main window so you can get the desired result.
To check how much we could adjust each rail we loaded this unit to its maximum of four times 15A on the 12V rails which is the maximum combined power. The 3.3V and 5V rails were also loaded to max. Running at its limit the rails all dropped below the ideal target values and we tried to use the software to regulate them back to normal. This resulted in the PSU switching off automatically. Remember: you can regulate the rails but there are limits to what you can accomplish.
Unfortunately, this kind of software doesn't come without problems. First of all we need to complain about the high CPU and memory load the software creates. With a typical midrange or better PC it would still be acceptable, but if you're running any demanding applications you might want to shut it off. Also, during the tests on the Chroma we connected the PSU to a second PC (which wasn't being powered by the PSU). While testing and changing to different tests the software always shut down and showed a connection problem with the PSU. This did not happen when the PSU and software were running in a normal PC configuration, but we did find it curious that the software apparently didn't like the results generated by Chroma.
Gigabyte has made a clear statement with the Odin series and we hope that there will be more in the future. The idea of providing software to control and monitor the power supply is very good and the results are generally worthwhile. The software itself needs to get a little tweaking here or there but already operates well at this point. The results shown sometimes don't match reality, in some cases they're worse than in others, but for the most part they are not too far off. The fan speed is very accurate though and also the shown voltage comes close to actual numbers. If the power output and amperage readings can be corrected, and hopefully reduce CPU and memory requirements a bit, the software would be about perfect.
Related to the above is the voltage regulation provided by the software. We didn't see any chance for this PSU after recording the bad voltage drops with increasing loads, but we were surprised at the ability of the software and hardware to regulate the rails, and the results were very good. Every rail had a better distribution after tuning, even at higher loads. However, there are limits to what can be done and when we tried to adjust the voltages with the PSU running at maximum capacity we found the limits of what could be done.
The efficiency of the Odin 800 is very high - higher than what we've seen from any previous CWT-models. At 85% with 230VAC input the result is very good and comes close to the Silencer 750 from PCP&C. Even better is that the overall efficiency is very good as well, showing a constant result of over 80% at any load higher than 20%.
The OCP has worked just fine on the 12V rails and unfortunately we couldn't figure out why it didn't work on the lower rails of 3.3 and 5V. The 24-25 amps on the 12V rails are a normal amount and used by most of the manufacturers which are rating the rails around 20A.
A potential concern is that the fan is not particularly quiet and never runs at a speed that could be considered "silent". You can clearly hear it when the power supply is running beside you. When mounted in a case with all the covers in place, it becomes much more difficult to hear, and as that's the normal usage scenario it's not a bad PSU. If you're looking for a quiet, high-efficiency PSU, the Odin 800 will fit the bill. If you're willing to risk overheating, you can even modify the fan speed to help keep noise levels low at higher temperatures, though we would be careful if you go that route.
At present, this power supply goes for around $230 USD and it's not yet listed in any European shops. There are power supplies that are available for less money that can provide equal or better results than the Odin 800, but it does provide some extras that might tip the balance back in its favor. The software that comes with this PSU is clearly worth taking into consideration, and you do get extra thermal diodes and a controllable fan header that some might find desirable. With any new technology there is always the potential for a high failure rate, but our initial results are positive and we hope that the build quality from CWT holds what it promised.