by Doors4ever
04/07/2009 | 05:36 PM
For many years I only dealt with Intel processors on testbeds, while my home systems were all built exclusively on AMD CPUs. Everything started long time ago with an AMD Athlon 750 Slot A CPU. There was a processor die in the middle of a PCB with two external cache-memory chips on both sides of it. This whole thing was placed into a humongous cartridge the size of a graphics card, only much thicker because of the cooling system. Then I stayed for quite some time with a renowned Socket A platform, a little bit with Socket 754 and then for a while again with Socket 939. What didn’t I like about Intel processors? There are, in fact, several different reasons. I didn’t like the constantly changing chipsets and sockets, when you had to replace the mainboard when you wanted to upgrade to a CPU with higher clock speed. For example, there were three types of Socket 370. I didn’t like Rambus memory, Pentium 4 processor microarchitecture and high heat dissipation of the Prescott cores. I didn’t like very confusing model lineup evolution, when the first Pentium 4 Willamette CPUs were slower than Pentium III, and the processors on the new Prescott core fell behind those on Northwood. Moreover, I wasn’t happy about the high prices on Intel processors that is why I always went for AMD CPUs that seemed to work perfectly for me from all aspects.
Of course, it would be incorrect to claim that I paid no attention whatsoever to Intel CPUs. However, I could totally satisfy my personal curiosity in the lab, where I could see their advantages as well as drawbacks in action. And I didn’t have any intention of switching over to Intel platform until things turned around in summer of 2006 when Intel announced their new processor microarchitecture implemented in Conroe. Spring of 2007 was the last time I tested a few Socket AM2 mainboards and purchased my very first Intel Core 2 Duo processor. After that I forgot about AMD platform for two long years. I don’t see the point investing into a platform that is guaranteed to be slower than the competitor even after overclocking. However, it would be unfair to say that nothing changed. AMD processors were now made with 65nm process, but ironically they followed Intel’s footsteps when the new Brisbane CPUs turned out slower than the old Windsor ones. The next milestone was the long-anticipated and still very disappointing AMD Phenom launch: Intel CPUs remained unattainably ahead. Time went on, priced continued to drop and finally we welcomed the new AMD Phenom II.
I would like to remind you that we have the whole bunch or materials on our site related to the new AMD solutions launched this year:
The review titles suggest that situation has turned around for AMD, although there were no revolutionary changes introduced in the new AMD processors. Intel CPUs still hold the performance leadership, however, the users have to put up with a lot of negative consequences resulting from this impressive speed, such as high power consumption, heat dissipation and extremely high price. Top processor and graphics card models have become mostly expensive toys for editors and dedicated enthusiasts. The developers use their flagship solutions to demonstrate their technological and intellectual superiority, while the regular users who follow the reviews of all flagship solutions with great interest in the end purchase absolutely different more mainstream components. This is where things have changed recently: nothing drastically new, but all the little changes combined together produced a serious breakthrough for AMD – their CPUs got in the spotlight of public interest again. The transition to new manufacturing process, microarchitectural changes, improved overclocking potential and reasonable price made AMD Phenom II processors serious competitors to junior models, i.e. more mainstream dual- as well as quad-core Intel CPUs.
We don’t know how long it is going to last, as Intel is preparing an extremely promising LGA1156 platform already, prices may change as well. So let’s take the advantage of the present moment and get back to talking about AMD platform. Today we are going to use AMD Phenom II X4 810 CPU to investigate the features and functionality of the new Socket AM3 mainboard from Asus – M4A78T-E based on AMD 790GX chipset.
According to the color scheme of the company logo and unlike the packaging for Asus mainboards for Intel processors, Asus M4A78T-E ships in a green box:
Both, the front and the back of the box are covered with logos of supported technologies and functions:
The accessories bundled with Asus M4A78T-E are pretty scarce and include the following items:
In our recent Asus P6T mainboard review we mentioned that every mainboard usually has a distinguishing feature. The funny thing is that Asus M4A78T-E seems to have no unique features: it is a good board, with pretty standard features and functionality. It is, in fact, also good. Most importantly, there are no evident drawbacks.
Asus M4A78T-E mainboard uses 8-phase processor voltage regulator circuitry that as always consist of very high-quality components. There is an individual phase for powering the integrated memory controller and HyperTransport bus. The power connectors are located in pretty common and therefore convenient spots. The only thing that surprised us was the 4-pin ATX12V connector instead of the 8-pin one although the board is claimed to support up to 140W CPUs. There is no formal connection between the processor power consumption and the number of pins in the ATX12V power connector, however, there is normally an 8-pin connector in this case.
Socket AM3 supports all CPUs designed in this form-factor. At this time AMD offers the following CPUs for Socket AM3: AMD Phenom II X4 and AMD Phenom II X3. However, Asus claims that this board will also support AMD Athlon X4, Athlon X3 and Athlon X2 CPUs, when (and if) they come out. Just in case, let me remind you that Socket AM3 processors have a universal memory controller that supports DDR3 as well as DDR2 SDRAM. Therefore, you can also install a CPU like that into a Socket AM2+ mainboard, but not the other way around. Asus M4A78T-E mainboard is equipped with four DDR3 DIMM slots that support up to 16GB of RAM total. In the nominal CPU mode the memory may function at 1066, 1333 or 1600MHz frequency.
The processor voltage regulator transistors and the chipset North Bridge are covered with relatively large heatsinks with sophisticatedly twisted fins. The chipset South Bridge heatsink is pretty nominal.
Asus M4A78T-E mainboard uses AMD 790GX North Bridge with ATI Radeon HD 3300 graphics core. As for the video memory, up to 512MB of RAM can be used for that purpose. Besides, there are also 128MB of memory on this board that have been assigned to the graphics subsystem via SidePort Memory technology.
When we look at the bottom part of the board, we can notice a COM connector that has been placed unusually high, right next to the chipset North Bridge heatsink. The graphics card slots have been moved a little farther from one another to allow easy use of graphics accelerators with massive cooling systems. If there is only one graphics card installed, it will work at full PCI Express 2.0 x16 speed. With two graphics cards used individually or together as an ATI CrossFireX configuration the slots will switch to x8 mode. The integrated graphics core can also be used together with a discrete graphics card due to ATI Hybrid CrossFireX support.
SB750 South Bridge delivers two PATA channels and six Serial ATA ports supporting RAID 0, 1, 0+1 and 5. Asus M4A78T-E is equipped with five actual SATA ports with the sixth laid out as an eSATA on the back panel. There are also 6 USB 2.0 ports out of the 12 supported by the South Bridge. The remaining six USB 2.0 ports are tied up to three pin-connectors along the lower edge of the PCB. There is an IEEE 1394 port next to them, which is implemented through VIA VT6315N controller. The second IEEE 1394 port provided through this controller is on the back panel. Interestingly enough, the eight-channel sound is implemented through another VIA controller – the VT1708S. I didn’t even known until today that VIA makes audio codecs, too. I have never seen them on mainboards until today.
We didn’t mention the PS/2 keyboard connector, six analogue audio-jacks and digital optical S/PDIF that are also on the connector panel of the board. Moreover, the integrated graphics core provides support for D-Sub, DVI and HDMI, while the Gigabit Atheros L1E controller delivers network RJ45 port.
The components layout scheme from the user’s manual will provide a better idea of Asus M4A78T-E design peculiarities:
We are going to wind up this part of our review with the detailed technical specifications of the Asus M4A78T-E mainboard taken from the manufacturer official web-site:
The BIOS of Asus M4A78T-E mainboard differs really slightly from the BIOS’s of other Asus mainboards we have reviewed before. So, it is going to be really easy to find your way through all these options and settings.
“Ai Tweaker” section that contains almost all the overclocking-related settings, by default doesn’t require the user’s attention: all parameters will be set automatically.
However, you can really appreciate the convenience of this section when you get to change some of the settings. All frequencies are reported as they actually are and they will change depending on the changes in the clock generator frequency. Only the actual values will be reported. The dangerously high values will be highlighted purple and then red.
The maximum processor Vcore can be set at 1.45V. If you want to be able to raise the CPU core voltage up to 1.65V, you need to reset the “CPU overvoltage” jumper on the board. This is more than enough to successfully overclock the CPU with traditional air cooling. However, if you plan on using more extreme cooling methods, you may need more than that.
The memory controller settings are singled out into a separate sub-section called “DRAM Controller configuration”.
The numerous parameters related to memory timings are also summed up in a separate sub-section.
The “Advanced” section has a sub-section with processor related settings. It would be more convenient to have its contents moved to “Ai Tweaker” section, where all important parameters necessary for complete system configuration are located.
The same “Advanced” section contains a subsection with parameters dealing with the integrated graphics core.
The next section called “Power” contains a very interesting subsection with monitoring options.
We are already familiar with the options in the “Tools” section. The built-in ASUS EZ Flash 2 utility for updating the BIOS became even more convenient to work with due to new ability to read from the HDD sections formatted for NTFS file system. “Express Gate” function allows booting rapidly a Linux based OS that will offer basic functionality, such as Internet access, instant messenger, photo viewing and editing, even some games. “AI NET 2” subsection will check the network cable status and “ASUS O.C. Profile can save complete settings profiles for easy access as needed.
However, they made a few changes even here. Unlike Asus mainboards for Intel platform, where you can only save two settings profiles in the memory, which is often not enough, Asus M4A78T-E mainboard allows saving up to eight profiles. Each profile can be given its own name reminding you of the settings it contains.
Overall, Asus M4A78T-E BIOS is very functional, very easy to work with. We only wish it had richer monitoring functionality.
Unfortunately, our first encounter with the Asus M4A78T-E mainboard was overshadowed by a number of different problems that have seriously affected our first favorable impression from it. It was later that we found out the board was not to blame for the discovered issues. However, it was not free from a number of its own drawbacks, which unfortunately surfaced later, too.
The main problem was the board refused to work stably even with default settings, without any CPU overclocking. The OS would boot, but any tests we tried running would inevitably lead to freezing and BSOD. For the preliminary tests I used the old operating system that was left from the previous LGA1366 platform tests. So, I decided to remove it and install fresh, and again, without any luck. Windows Vista installation completed successfully, but everything would freeze upon first system boot-up when the performance check is made.
Finally, it turned out that there was a compatibility issue between the board and 1024MB DDR3 OCZ PC3-14400 Platinum Series modules. I had two OCZ3P18002GK kits at my disposal, so the number of DIMMs was not the issue. I tried four, two and even one and checked out all the DIMM slots. The system remained unstable or couldn’t boot Windows Vista at all; couldn’t save the BIOS settings saying that it detected a new CPU on every system restart.
At that point it was unclear who the one to blame was: Asus or OCZ? Therefore, I continued the experiments on Asus M4A78T-E with different memory modules. However, when I got to the preliminary stability tests, we put together another Socket AM3 testbed based on Gigabyte GA-MA790XT-UD4P mainboard. This solution is based on AMD 790X chipset and its review will come out shortly. At this point I would only like to say that the Gigabyte board also refused to work with DDR3 OCZ PC3-14400 Platinum Series memory modules. So, does it mean that it’s all OCZ’s fault? Any maybe we should blame AMD, since their memory controller is integrated into AMD Phenom II X4 810 processor?
They say that it happens very often that mainboards are incompatible with memory some modules, video cards or other expansion cards and it doesn’t really depend on the platform type. It can happen to an Intel platform as well as to an AMD one. I personally have tested dozens of mainboards, but cannot recall any compatibility issues like that in my experience. For the first time in two years I got to testing a mainboard for AMD processors and immediately got a compatibility problem. Two different mainboards, from two different makers, based on different (though related) chipsets, one has an AMI based BIOS, while the other – a Phoenix-Award based one, and both of them do not work with the same memory modules. I still have my doubts about the guilty, but for some reason I just wanted to get back to testing Intel platforms again…
I replaced the memory modules, but it didn’t solve the problem. When I started my overclocking experiments and found what seemed to be an optimal and operational combination of frequencies and voltages, another mysterious phenomenon surfaced. The system successfully passed short stability tests with utilities creating high workload, but would fail any long-term tests lasting for an hour or more. Everything would inevitably lead to the notorious blue screen of death. And again it took me a while to determine that it wasn’t the mainboard’s fault, but the power supply’s. By pure coincidence or irony the PSU was an OCZ one, too.
I still do not say that OCZ solutions have a quality issue. However, I can and will complain that the fan inside my OCZ GameXStream OCZGXS700 PSU was so loud that I couldn’t use it in my home system and had to replace with a quieter and hence less powerful one. This modified power supply unit proved capable of working just fine in my moderately overclocked home system for a long time. Moreover, it successfully participated in our recent tests of an overclocked and hence very resource-hungry LGA1366 platform. However, the quiet but less powerful fan could no longer cope with the overheating PSU during long-term stability tests. That was the reason for the blue screens of death to pop up.
Finally, when all the problem-causing parts have been successfully identified and replaced I tested Asus M4A78T-E mainboard in the following platform:
We used Microsoft Windows Vista Ultimate SP1 x86 operating system and ATI Catalyst 9.2 graphics card driver. It turned out that if you run a 32-bit OS and use 4GB of RAM, you can increase the amount of memory assigned to the integrated graphics core “for free”. 32-bit Windows Vista can only use 3.3GB of memory, so we could provide our Radeon HD 3300 with the maximum possible amount of 512MB without losing anything. The OS will still have all the 3.3GB it can work with. Of course, it doesn’t necessarily mean that the performance of the integrated graphics core will increase dramatically because of that, but it won’t do any harm, that’s for sure.
In fact it is pretty interesting to see how UMA and SidePort technologies actually coexist? Asus M4A78T-E mainboard has 1q28MB of DDR3 memory assigned for graphics subsystem needs through SidePort technology. It works at 1333MHz by default with unknown timings. The integrated graphics card can also use system RAM that can work at a completely different frequency and with different timings. So how do they get synchronized? Will the “excessive” requests the graphics card sends to RAM affect the system performance, can it potentially drop? We tested our system in its nominal mode in order to answer all these questions. In the first test session it was the integrated ATI Radeon HD 3300 graphics core that employed the 128MB of default SidePort memory along with 512MB from the system RAM. In the second round of tests there was a discrete Radeon HD 4870 512MB graphics card with disabled integrated graphics core.
Please disregard the specific numbers showing a tremendous performance difference between a powerful discrete card and a weak integrated graphics core. The tests took a long time but completed successfully, although 3DMark Vantage offered to change the test settings and rerun the test, because the results looked really suspicious.
The main conclusion is that the general system performance doesn’t get affected by the use of the integrated graphics core in those benchmarks where the graphics card has no influence on the final result. As for ATI Radeon HD 4870 and Radeon HD 3300, we all know that we can’t really compare the functionality and potential of the two, even though we use medium quality settings for both in all graphics tests.
I would like to say right away that Asus mainboards’ specifics that do not let us overclock Intel CPUs well and at the same time maintain all processor power-saving technologies, are not the case with AMD processors. You can lower the processor Vcore, increase it or leave at nominal and the CPU will work as instructed under load.
However, as soon as the load disappears AMD Cool’n’Quiet technology will kick in and the CPU will lower its frequency and core voltage.
Overall, I have been very please with AMD Phenom II X4 810 overclocking experience on Asus M4A78T-E. Overclocking hasn’t been so simple, clear, quick and easy for a while. Just raise the clock generator frequency and increase the processor Vcore if system loses stability monitoring the temperature at the same time. And then increase the frequency again. The only other thing you should keep an eye on is that the frequency of the North Bridge integrated into the CPU and the HyperTransport bus connecting the processor with the chipset North Bridge were close to the nominal 2000MHz. Taking into account the potential of Asus M4A78T-E mainboard, all these requirements are really easy to meet. For example, if I am ever invited to my kids’ school for an overclocking lesson, I will take something close to my current testbed configuration rather than an LGA775 or LGA1366 platform, which have much more things to take into account.
Our AMD Phenom II X4 810 processor at 230MHz frequency passed preliminary stability tests in LinX utility on Asus M4A78T-E mainboard without changing its Vcore from the nominal 1.3V. For your reference, the same processor managed to pass the same tests on Gigabyte GA-MA790XT-UD4P mainboard at 250MHz. Quite a difference, isn’t it? However, I didn’t know about it at that time. Besides, AMD processors do not require nominal Vcore settings to be in place for processor power-saving Cool’n’Quiet technology to work. Therefore, we continued our experiments. The end result turned out pretty good: we managed to overclock our CPU to 285MHz base frequency, i.e. 3.7GHz clock speed. Besides LinX utility, we tested the system stability with more than an hour run of Prime95 in Blend mode.
To increase the system performance we raised the frequency of the North Bridge integrated into the CPU to 2565MHz. that in its turn required raising CPU/NB Voltage to 1.275V. The processor core voltage had to be increased to 1.575V, but in idle mode it would drop together with the frequency multiplier.
By the way, when we used the integrated graphics core we had to stop at 275MHz frequency, because further increase led to evident system instability and serious image quality artifacts. It seems that maximum overclocking of a system using integrated graphics will seriously depend on the current memory frequency and timings and on the integrated graphics core settings in the mainboard BIOS.
When we finished this round of tests on Asus M4A78T-E mainboard, we put together a testbed around Gigabyte GA-MA790XT-UD4P in order to check if OCZ memory would work ok and see how far we could push our test processor here. As you know, the memory wouldn’t work, CPU overclocked way higher at the default processor core voltage, but the results of maximum CPU overclocking test remained exactly the same. We had to increase the CPU Vcore only to 1.525V, while the additional +0.15V had to be provided for the North Bridge controller integrated into the CPU.
We are going to discuss Gigabyte GA-MA790XT-UD4P in detail in our next review, and now at the final testing stage we are going back to our today’s hero - Asus M4A78T-E mainboard. We reinstalled the operating system and ran performance tests in the nominal mode as described in the previous part of the article. After that we launch the same set of benchmarks on an overclocked system, but suddenly the BSOD comes up during 3DMark Vantage run. It was very strange, because we reproduced exactly the same testing conditions as before. Tried to change a few things, but again no luck. Desperately looking for answers in the preliminary test results obtained on Gigabyte mainboard, and notice that it required higher voltage on the North Bridge integrated into the CPU to ensure stability. So, we increase this setting a little, then another bit and finally pass the notorious test. And things could have been great, however, the system could no longer pass the long-term stability check in Prime95.
By that time I was already a little exhausted by all the problems I faced since the beginning of the Asus M4A78T-E test session. However, I reduced the overclocking to a completely different reason. Asus M4A78T-E mainboard can send maximum 1.45V to the CPU by default. In order to be able to raise the Vcore to 1.65V, you have to reset the “CPU overvoltage” jumper on the board. We did this exact thing to ensure that we could overclock our processor to its maximum. However, we started to get “Overvoltage error” message pretty frequently on system boot-up. It required pressing F1 to continue booting. I didn’t pay too much attention to it for a while, but then I went into the Hardware Monitoring section. Although the system was working in its nominal mode and I didn’t raise any voltages, the CPU was getting over 1.64V! I immediately recalled that Everest often reported higher CPU Vcore when we ran the tests on an overclocked processor. Namely, it read up to 1.62-1.64V, although we had only increased it to 1.575V in the BIOS. Check out a couple of screenshots above. I assumed that it was a monitoring error, but turns out that the CPU was in fact receiving that much voltage. Could it be slowly killing our CPU thus affecting its overclocking potential? Well, we will find out during our Gigabyte board tests, and at this point we’d better give up “CPU overvoltage” function altogether.
I usually record my experiments in a special overclocker log. That is why I immediately remembered that at 1.45V processor core voltage the system passed LinX stability test at 270MHz clock generator frequency, but refused to restart. It would freeze in the very beginning, during the graphics card initialization process. At that time I thought that it happened because of overly overclocked processor and used “CPU overvoltage” jumper to raise its Vcore a little further. This time let’s try and lower the clock generator frequency to 265MHz – no restart. The same happened at 260MHz and only when we got down to 257MHz we could reboot the system successfully. However, this is relatively low frequency. Gigabyte mainboard could reach just a little lower frequency even with the nominal processor Vcore…
Well, there is a lot of stuff connected with the processor core voltage here, so I decided to try and lower it a little and see what happens. This was the right thing to do. Now the board could reboot at 258MHz clock generator frequency. After a couple of additional experiments I discovered that when the CPU Vcore is lowered down to 1.4V, Asus M4A78T-E mainboard can even restart successfully at 270MHz. That was almost a victory! I say “almost” because this voltage was not enough for the system to pass the tests at this frequency.
Finally, when the processor Vcore was at 1.4V, Asus M4A78T-E mainboard could only ensure satisfactory operational conditions for it at 250MHz. this is where we had to stop.
Actually, 250MHz frequency is very convenient for CPU overclocking. In this case we get nominal frequencies for the memory, North Bridge integrated into the CPU and HyperTransport bus. In fact, we increased the North Bridge frequency to 2500MHz and lowered the memory timings in order to pump up the speed.
Now we need to find a worthy rival for our test processor. From our review called “Meet Socket AM3: AMD Phenom II X4 810 CPU Review” we know that this processor can successfully compete against Intel Core 2 Quad Q8xxx series, however, I didn’t have a CPU like that at my disposal at the time of tests, unfortunately. Instead I had the new Intel Core 2 Quad Q9400 with yet unknown overclocking potential. Would you like to check it out now?
Luckily for us and unfortunately for the undertaken comparison, the CPU overclocked brilliantly. So far, I have never exceeded 500MHz FSB in my overclocking experiments on quad-core Intel Core 2 Quad processors.
Sadly, even when if overclocked AMD Phenom II X4 810 to 3.7GHz, it wouldn’t stand a chance against the overclocking monster from Intel. So, our 3.25GHz achievement is absolutely out of the question. Well, since we don’t have a fair rival at this point, we will just stick to on single CPU in the next round. It will give us another change to check out the advantages from CPU overclocking.
Not bad at all. We overclocked our CPU by 25% and got the same performance boost. Of course, lowering the memory timings and increasing the integrated North Bridge frequency helped achieve these results.
We measured the power consumption of our test systems running in the nominal mode with a discrete graphics card as well as integrated graphics core and in an overclocked mode with a discrete ATI Radeon HD 4870 graphics accelerator. We used Extech Power Analyzer 380803 device. This device is connected before the system PSU, i.e. it measures the power consumption of the entire system without the monitor, including the power losses that occur in the PSU itself. When we took the power readings in idle mode, the system was completely idle: there were even no requests sent to the hard drive at that time. We used Fritz Chess Benchmark to load the CPU and FurMark utility to load the graphics subsystem. We recorded the maximum readings.
Of course, CPU overclocking by raising the core voltage increases the system power consumption. However, in this case the transition from a power-efficient integrated to a powerful discrete graphics solution has a much bigger influence on the end result.
At first, numerous issues that we faced throughout the en tire test session wouldn’t let us adequately experience the benefits of the Asus M4A78T-E mainboard. Luckily, it turned out that the mainboard has nothing to do with most of these issues, and is overall a very good product. Asus M4A78T-E boasts very convenient design and sufficient functionality. It will be a good basis for a power-efficient and very quiet system if used with an integrated graphics card, or high-performance system if used with a powerful discrete one. There are only two issues that prevent the board from reaching the top of its ability:
If Asus engineers manage to eliminate these issues in their upcoming BIOS versions, we will have an excellent, almost universal mainboard. If not, then Asus M4A78T-E will just remain a good board, working stably in nominal mode, but capable of modest CPU overclocking.