An intensive look finds a high quality product – Brett Wasserman
Today at Overclockers.com we have one of the newest Intel 845PE offerings from a not so well known Taiwanese company called Albatron. While Albatron has been around since 1984 making various display products, it’s only in the past few months that they entered the cut-throat motherboard and video card markets.
On the test bench for review this time is the Albatron PX845PE Pro II. Since I’m sure all of you have read numerous 845 motherboard reviews, I’ll skip the usual “cut and paste” of the manufacturer’s specs and refer you directly to Albatron’s product page for them instead.
As I mentioned in my last review of the Iwill P4HT, we will be using a non-traditional but very informative series of tests to shed some light into the strengths and weaknesses of various motherboards.
In order to methodically stress test, we use a pair of PCI based hardware diagnostic cards manufactured by Ultra-X. One of them is the RAM Stress Card and the other is the P.H.D. PCI tester. These cards require no operating system to function and will clearly identify the exact mode of failure when we overclock a motherboard to the breaking point.
Our testing philosophy for this series of reviews is to use an unlocked CPU and very high quality RAM at very relaxed timings. This should allow us to overclock the FSB of the motherboard under test while doing our best to eliminate the CPU and the RAM as possible points of failure. Using some electronic test equipment, I’ll also “show” some of the important power related signals on the motherboard while it’s being stressed.
Our test rig for this review consists of:
Albatron PX845PE Pro II motherboard (silver version) with –
- Unlocked Pentium 4 1.7 GHz Willamette CPU
- 256M Corsair PC3500 CAS2–one single bank module
- Vantec 470 watt Stealth power supply
- Liebert 2.7kVA uninterruptible power source (UPS) with integrated power conditioner
- Matrox G550 AGP video card
- Vantec Aeroflow heat sink fan assembly with 5800 rpm TMD fan
- 2 YS Tech 80mm speed adjustable 12Volt fans
- Floppy drive (for loading test script and saving results)
The BIOS in the Albatron motherboard, which incidentally is a dual BIOS, just like the Gigabyte boards have, comes from the familiar folks at Award/Phoenix. This is an easy BIOS to work with and has all of the necessary overclocking features that you might want. No particular shortcomings are apparent here.
Above we see the single screen related to overclocking. FSB frequency adjustments from 100 to 248 MHz are available in 1 MHz increments. Vcore is adjustable between 1.10 vdc and 1.85 vdc in .025 volt increments. Vdimm can go from 2.5 vdc to 2.8 vdc in 0.1-volt increments, and for those of you who need to pump more AGP voltage, 2.5 vdc and 2.6 vdc are available. DDR/CPU ratios available are 1.5x, 2x, 2.5x, Auto and 2.66x.
As with most other Intel chipset based motherboards, we also have the ability to lock the PCI and AGP clocks at 33 and 66 MHz respectively.
Next we can see the BIOS advanced settings screen. Here we find the typical settings for adjusting memory options. Available settings are:
- CAS Latency: 1.5, 2, 2.5 cycles
- Active to Precharge Delay: 7, 6, 5 cycles
- DRAM RAS to CAS Delay: 3, 2 cycles
- DRAM RAS Precharge: 3, 2 cycles
- Refresh Mode: 7.8, 15.6, Auto, 64 microseconds
Our first test, as always, is to run the motherboard with all of the components at their defaults:
- CPU = 17 * 100
- Vcore = 1.75vdc (default for Willamette CPU)
- Memory by SPD
- Memory ratio = Auto (2.66x=DDR266)
- PCI/AGP locked at 33/66
With these settings, I ran each of the Ultra-X cards (PCI tester and RAM stress tester) for 2 hours. Absolutely no errors were encountered. Below you can see the group of tests that I’ve configured the PCI tester to run. (In the current version of the PCI tester’s firmware we are now able to run all of the RAM stress tests as well, in our test suite.)
So far so good…Now on to raising the FSB and lowering everything else.
In the course of doing this review, I discovered a very interesting little tidbit about our unlocked Pentium 4 Willamette chip with respect to the multiplier: The default multiplier for this 1.7 GHz CPU is 17. If I install it in a motherboard and go into the BIOS setting for the multiplier the lowest setting, I can see is 10X.
If I enter anything below 10, whatever was entered initially remains. I can enter any value between 10 and 24. If I clear the CMOS using the jumper, the default returns to 17. So, it would appear that the lowest multiplier that I can use is 10X.
But, I discovered that if I set the BIOS settings to a point where the motherboard will no longer boot and the watchdog timer causes a reset, the value of the multiplier defaults to 8X.
While you normally wouldn’t be looking for a way to further UNDERCLOCK your CPU, this technique let’s us further limit the CPU as the overclocking weak point. If all other things are equal (sure, right, hahaha), this lower multiplier should let us reach a higher FSB setting on motherboards we are testing.
For the overclocking stress tests, we start with these settings:
- CPU = 8 * 170
- Vcore, Vdimm, VAGP=default
- Memory settings = 2.5, 7, 3, 3
- Memory ratio = 1.5x (=DDR255)
- PCI/AGP locked at 33/66
Again, we see no failures with these settings. Ramping up the FSB while keeping all other settings the same, we begin to see errors at an astonishing FSB setting of 220 MHz. The highest FSB setting, with the other settings as above, that would run without errors reported by the Ultra-X cards was 219 MHz. Just to be certain that this wasn’t an anomaly, I ran each of the test cards for at least 12 hours at 219 MHz and again, no errors were reported.
When the RAM stress card reported errors at the 220 MHz FSB setting, it very clearly told us that the failure mode was the inability to maintain integrity in reading and writing to RAM chips 2 and 7. Extended testing at 220 MHz also revealed that RAM chip 3 also had difficulty being read from and written to, but with far fewer problems.
Some of you might be wondering if the errors reported above are indicating a RAM problem rather than a failure in the motherboard. If we were specifically testing the RAM (as the only unknown), and we KNEW that motherboard and CPU would operate at the given settings, then we would be seeing a RAM failure.
But, since we KNOW that our Corsair DIMM is rated for DDR433 and that we are only operating at DDR330 (220FSB * 1.5 multiplier), we can feel confident that the indicated failure is related to signal timing or driving of the memory bus, and not the RAM itself.
Now that we know where the breaking point is for this motherboard (219 MHz = good, 220 MHz = fail), let’s take out the oscilloscope and take a look at the important voltages being delivered to and generated by the motherboard. This should give us an indication of the physical quality of the board and some of the components.
Above we see a totally noiseless and clean 3.3-volt line coming in from the Vantec 470 watt Stealth PSU. All three of the important power supply voltages (+5v, +12v and +3.3v) look identical to this scope trace. Very, VERY nice indeed.
As with most (but not all) Pentium 4 motherboards, a three-phase synchronous buck circuit is used to take 12 volts from the PSU and “buck” it down to the voltage that we need for Vcore. A very detailed discussion of this circuit, along with a schematic can be found HERE.
Here we see the output side of one of the MOSFET pairs:
This is the pulse that is delivering power to the CPU one third of the time.
Unlike past motherboards that I’ve seen, this Albatron board has surprisingly little overshoot at the phase intersections (the times that the MOSFETS pairs switch on and off). In addition, the duration of the overshoot is VERY short. While I didn’t include a scope shot of it, the duration of the voltage spikes shown is so short (less than 2ns) that I was not able to get a good measurement of them even with the scope set to display at 10x time resolution (2ns per screen division).
Next we take a look at the same phase of the circuit after being combined with the other 2, but before the big capacitors get a chance to smooth out what will become the Vcore for the CPU:
Here things look lousy, but this is because the circuit is picking up ringing and overshoot from all three phases. If this signal were presented to the CPU we would have numerous problems, if we were able to get the system to run at all.
After the capacitors smooth it out:
Ah…that’s better. Nice, clean, boring horizontal straight line. Exactly what we want to see going to the CPU.
Last week we received a very useful new piece of equipment here at Overclockers.com. Let’s take a look at what we can do with a DC current clamp probe.
Joe has been bugging me to come up with a way to measure the current being drawn by a motherboard under stress. Knowing the amount of current draw can give us some insight into quite a few things that many of us have speculated about for some time. How big of a power supply does a given system REALLY need? How efficient is our power supply?
A current clamp probe is a very easy device to use. Unlike typical voltage probes, a current clamp probe does not have to make physical contact with the conductors (wires) carrying the current. The conductors carrying the current in the positive direction towards the motherboard emit a magnetic field proportional to the amount of current.
The current clamp, if positioned properly with only the wires relating to one voltage in between the jaws, will react to the magnetic field using the Hall Effect (explained HERE.) If you look closely at the picture, you’ll see that the wires from the PSU are separated and bundled into their respective colors.
When we measure current, we need to place the clamp around ONLY wires of the same color. This lets us isolate the current flowing from each voltage supply rail. Our current clamp outputs 1mv per DC amp, and we connect it to our DVM for display.
Current Measured – Amps
|12vdc (at ATX connector)||
|12vdc (at P4 connector)||
These current values are quite interesting. What we see is that the 3.3v and the 12v supplies are the most important in running the motherboard. Very little current is being drawn from the 5v rail.
And, for Joe’s benefit, we see that the total power dissipation from this group of components (motherboard, CPU, 3 fans, DIMM, keyboard, mouse, 1 PCI card, 1 AGP video card and a floppy drive) under high stress, is only ~71 watts. Looking at this in the extreme, if the total efficiency of the overall system was only 10% (and it’s certainly higher than that), that would mean that there would be 64 watts of heat being generated.
We can also surmise that the total heat dissipated (theoretical limit at 0% efficiency) by the combination of components being fed by the separate 12v connector wires is ~49 watts. That particular point might give Joe some greater insight to the actual cooling requirements of Pentium 4’s in general.
Looking at these numbers from a more general level, we see that power consumption is low even under extreme conditions. Even if we were to fully load up the PCI slots with additional cards and install a power hungry graphics card like the Radeon 9700 Pro (with it’s extra power connector), it’s likely that we could not come close to doubling the current draw.
Among other things, it looks like we just confirmed that the minimum 300 watt power supply that Intel specifies is MORE THAN SUFFICIENT for even a loaded system.
Measured with DVM
The layout of the board and the overall build quality is OH SO CLEAN.
In the olden days when I used to design hardware, circuit boards were laid out in a very clear, but flexible, grid pattern — much like a newspaper page. Most of today’s motherboards, though, are laid out in an organic manner — from the major components outwards.
Albatron has done an excellent job of combining these two layout approaches to achieve both a convenient and orderly design, which certainly contributes to the board’s excellent performance. The only slight negative here is that the two USB headers for channels 5 and 6 are located right between the AGP slot and the first PCI slot.
The silver color of the board is quite nice to look at – it would look great in a clear case. I’ve had my doubts about the silver coating on this and some other boards on the market (Soyo and Triplex come to mind) being anything other than an esthetic feature. The manufacturers claim that the coating reduces EMI and dissipates heat better.
Although I can’t comment on the latter, I can say that I had an easier time making measurements with the oscilloscope than with any other motherboard that I’ve worked with. (I won’t get into issues like scope grounding, probe types, etc., but suffice to say that the signals were just “cleaner” and easier to trigger on. Maybe there really is something to this silver coating – maybe not. If someone has an emi spectrum analyzer that they want to loan us, maybe we can find out.)
The highest overclock that I could reach with a stock (locked) Pentium 4 2.53B GHz CPU, and using the Ultra-X cards to perform testing, was:
- FSB=164 (164 * 19 = 3116 MHz)
- Vcore = 1.55V
- Memory = DDR436, CAS2.0, 7,2,2
- VDIMM = 2.6v
- Cooling = Vantec Aeroflow with 5800 rpm TMD fan
The one very little quirk that I noticed with this motherboard was that it never reacts to my first key press when starting up and when entering the BIOS. No matter what I did, the board required 2 presses of the delete key to get into the BIOS, and then 2 presses of the arrow keys to begin to navigate within the BIOS.
As I mentioned earlier, this board has a mirrored BIOS. I simply love this feature and feel that it should be standard on enthusiast-targeted motherboards.
Unlike the Asus POST reporter that speaks even when everything is fine, Albatron wisely chose to only have the board speak when something is actually wrong. One simple, standard beep tells you that everything is OK.
To sum up, I REALLY like this motherboard.
It works great and is an excellent overclocker. The included features are well chosen and only lack on-board Firewire support. (Just so that you are aware, Albatron did not send us this board for review. I purchased and paid for it myself. While we did request a contact at Albatron to answer some questions, they did not respond to our request. Hopefully, this isn’t representative of their handling of customer support and RMA issues.)
This board is going to end up in one of the systems that I use routinely. I’m simply not aware of anything that Albatron could have done to make this an even better motherboard. I hope that they continue to produce great products like this.
Next up on the review block, due to popular demand, at Overclockers.com will be the SIS 648 offering from the folks at Asus — the Asus P4S8X. The good guys at Directron kindly are lending us this board to test. Keep your eyes open for it.