 |
Printed On: | May 10, 2008, 3:20 am |
| Category: | Chipsets & Motherboards | |
| Type: | Reviews | |
| Posted By: | Firefox | |
| Date Posted: | March 26, 2008, 10:34 pm |
 |
GIGABYTE EP35-DS4
It seems that going Green is the latest fad amongst motherboard manufacturers. With ASUS launching their motherboards with their EPU and GIGABYTE revamping their entire line of P35 motherboards with their Dynamic Energy Saver (DES) technology. Whilst ASUS' method is more traditional, adjusting the processor clockspeed and voltage to save energy, GIGABYTE has decided to take things one step further. In addition to the usual voltage and clockspeed adjustments, DES helps to save power further by shutting down power regulation phases when they are not required. However, both technologies are not that different as they both employ the PWM controller chip to go green. With ASUS, it's a customized controller from Analog Devices and with GIGABYTE, their engineers teamed up with Intersil to provide a software/ hardware solution.
If you have been tempted by the the Intel Quad-cores going for a song but are wary of the heat put out by these processors, you will surely appreciate the fact that GIGABYTE has revamped their entire line of P35 boards with DES technology. In fact, if you have a later revision GIGABYTE P35 board, a BIOS update is all you need to reap the benefits of DES. Whilst it may seem wierd that GIGABYTE has chosen to refresh boards based on a chipset that is nearly a year old, the fact is that the Bearlake proves to be an immensely popular chipset. It's inexpensive, offers great cost to performance ratio and for the enthusiasts, overclocks very well. Rather than pay a premium for the X38/ 48 chipsets and DDR3 memory, users can opt for a P35 board and still have enough spare change for performance DDR2 memory to push their processor. Today, we'll look at one of GIGABYTE's refreshed P35 motherboards - the EP35-DS4 - and determine if the DES feature does work and just how much it affects power savings in a system.
GIGABYTE claims up to 70% power savings and 20% better power efficiency!
Many users might feel that the board seems to be the same as the P35-DS4 and indeed, the board is essentially the same but for the BIOS/ firmware required to allow the DES software to interact with the board and some new bling bling related to DES. We will cover that in a moment so lets take a look at the basic specifications before moving on.
Basic Specifications
| CPU |
|
| Chipset |
|
| Memory |
|
| Audio |
|
| LAN |
|
| Expansion Slots |
|
| Form Factor |
|
| Storage Interface | South Bridge:
|
| IEEE 1394 |
|
| USB |
|
| Internal I/O Connectors |
|
| Back Panel Connectors |
|
| I/O Controller |
|
| H/W Monitoring |
|
| BIOS |
|
| Unique Features |
|
|
Pictures Galore
If it ain't broke, don't fix it! The board is almost identical to the tested and proven P35-DS4.
6-phase power is used on the board along with all solid-polymer capacitors.
EPS 8-pin power connector is used to feed more demanding CPUs
The 24-pin ATX connector is now a norm but the 4-pin molex is optional and is used only for CrossFire on the 2nd PCI-E x16 slot.
|
Continued.
A good mix of PCI-E x1, x16 and PCI slots allow for good expansion options!
6 SATA-II ports are from the ICH9R whilst the Gigabyte SATA2 (JMicron) controller as provides 2 more SATA-II ports in addition to the IDE port.
For those who need it, the Texas Instruments Firewire controller is a godsend whilst Gigabit connectivity is handled by a Realtek RTL8111B PCI-e network controller.
Aside from the usual manual and CDs, Gigabyte provides a SATA to e-SATA bracket along with an assortment of cables.
Note the mixture of right-angle and straight SATA cables!
|
Test Setup.
Test Setup
- Intel Core 2 Extreme QX6850 @ 3.0GHz (9 x 333MHz)
- Gigabyte EP35-DS4
- Aeneon 2GB Dual Kit DDR2-1066 CL5 @ 1066MHz, CL5-5-5-15
- Inno3D Geforce 9600GT 512MB (Reference clocks)
- 2 x 80GB Seagate Barracuda 7200.7 SATA HDD
- Pioneer DVR-215BK DVD-RW
- Enermax Galaxy 1kW PSU
- Intel Chipset Drivers 8.3.1.1009
- NVIDIA ForceWare 174.16 WHQL
- Windows Vista Ultimate x86
The GIGABYTE EP35-DS4 is based on the older P35-DS4 and so we will be looking at the performance impact (if any) from enabling the DES features rather than comparing it to other boards. The performance of boards based on the P35 chipset are a dime a dozen and the same goes for reviews of the P35-DS4 boards. Hence, we will concentrate on the DES feature of the EP35-DS4 instead.
|
Dynamic Energy Saver (DES)
GIGABYTE teamed up with Intersil to develop a software to control and monitor the DES feature on the board and this software also shows how much power is being consumed by the CPU even when you choose not to enable DES. Note that DES does not work when you're overclocking and the panel does not allow you to do so. We decided to test the DES feature under 2 conditions: Idle, and at Half-load. The latter was achieved by using one instance of Orthos (2 cores). These were run with DES enabled and disabled to monitor the differences in power consumed, if any. At the same time, we would also disable CPU throttling only to see how much power savings are to be had from just voltage throttling.
We also measured the power consumption at the wall of the system (sans LCD) to determine if there are still any power savings to be had after factoring in power supply inefficiencies. Obviously, for DES and any similar technology to really provide any savings, they would have to save more than just a handful of Watts at the board side. Without further adieu, the results:
Idle in Windows
Here, we can see that DES saves 15W of power but results in only a mere 1W to 2W of savings by the time PSU efficiency is factored in. Phases 5 & 6 have been shutdown to save some power in addition to clockspeed throttling.
Half-load with 1 Instance of Orthos Running
With 2 cores out of the 4 loaded, the software does not throttle the CPU voltage nor does it shut down any phases. Only when CPU throttling is enabled do we see any power savings. In this instance, it relates to 25W less CPU power consumed and 10W difference at the mains due to PSU efficiencies. Overall, this issue will only exist on quads, as it isn't too hard to load dual core processors with the software technology that is available today. Results may vary though, especially with the myriad of 80+ power supplies available in the market now.
In case you are wondering, we tested DES at full load conditions but found that there were absolutely no differences at all. Due to varying power supply efficiencies with load, the amount of power savings to be had varies largely depending on the setup employed. Obviously, the higher the efficiency of the power supply, the more the savings. If you're thinking that power saving is just dependent on the board and not the power supply, you would be very surprised. Don't skimp on a good power supply and you will see the savings in the long run. That said, there are additional benefits than just the power bill. With power saved on the board alone, you will also have less heat being emitted from the system. Being as heat is an enemy of electronics, one can also expect a more reliable system as well. Overall, a very nice move by GIGABYTE and Intersil.
Intersil's ISL6327 is in charge of the hardware side and LED's are used as a physical indication of the phases (red for OFF, green for ON)
|
Synthetic Benchmarks
To test the performance impact from DES CPU throttling, we ran a couple of synthetic system benchmarks including the new FutureMark PCMark Vantage as well as PCMark 05. Of course, we will not leave out the ever popular 3DMark 06 benchmark either or LightsMark 2007 which, you'll see, is as dependent on system performance as it is on graphics hardware.
In 3DMark 06, there is an impact of 510 points from CPU throttling.
PCMark 05 tests system performance and here, the CPU throttling results in a drastic 1500 point gap.
The latest PCMark - PCMark Vantage - The impact on the score is approximately 18%.
At 1280 x 1024, enabling DES results in a 20 FPS penalty in LightsMark 2007.
In SuperPi 8M, we see an extremely large performance impact. Clearly, throttling was at its maximum as the software only sees the CPU load as 25% (1 in 4 cores)
|
Games
We ran a couple of games to see how much of an impact would be seen in real world applications. Here, we have World in Conflict, Unreal Tournament 3 and Quake 4, all run at 1024 x 768 with low settings and with AA/ AF disabled. This would remove the graphics card as the bottleneck.
World in Conflict proves to be very dependent on CPU performance and we see a 28% penalty in framerates.
In Unreal Tournament 3, we see an insignificant difference of 2 FPS.
In Quake 4, we see a fairly huge gap of 55FPS - a 31% difference in framerates.
|
Conclusion
Conclusion
The move to go green is certainly a refreshing change from the usual performance/ overclocking rat race. With the Dynamic Energy Saver function, I found that the most power savings to be had was from shutting down the excessive power phases rather than just throttling the CPU voltage or operating frequency alone. With the larger motherboard makers pushing for ever more power phases on their motherboards, it's good to see that GIGABYTE/ Intersil have recognized that shutting down power phases that are not required would actually reduce power consumption significantly. This, even as many enthusiasts amongst us look to reduce the power consumption of auxiliary systems that are on 24/7 by using the typical methods: reducing frequency and voltage.
Unfortunately, not all is as rosy as it looks. The CPU throttling feature in DES has a significant impact on performance as the software seems to have issues detecting CPU load with multi-core processors. That is just about every processor on the market these days. When running 2 instances of Orthos to load the all the cores of the Kentsfield, the throttling was disabled almost instantaneously - an indication that the software only sees basic performance counters from the operating system rather than the load on each core. Hence, if you're running a single threaded application that fully utilizes one core out of a quad-core CPU, DES would not disable the throttling.
If you are looking at the bottom line and want to know whether DES is a failed attempt, I can safely say: No. As can be seen from the power consumption tests, CPU frequency throttling makes for insignificant power savings but it has a major impact on performance. However, the shutting down of phases does have a significant impact on the power draw and it does not compromise on performance. Hence, my advice for would-be users is to enable DES and voltage throttling but not to enable CPU frequency throttling. Also, it will be nice to see DES implemented on GIGABYTE boards since they are fairly feature rich these days and are potential home server candidates that will be left on 24/7 with lots of idling time.
Although the technology is still new and not mature at the moment, there are already promising results from the implementation of dynamic power phases. It's something new and unheard of prior to this and I hope that more can be done to polish up the software (CPU frequency throttling) in the near future. DES is certainly something interesting coming from GIGABYTE and should prove to be a yardstick for other manufacturers who are in the Green race.
|