A comprehensive look – Brett
Today on the test bench here at Overclockers.com we have the new Iwill P4HT, an 845PE based, high-end motherboard. This new release from the kind folks at Iwill has tons of bells and whistles (LED chipset fan, RAID, 6 channel audio and much more). It’s also RED!!! Everything about this board screams out fire engine RED!!!
Rather than simply describe the board and run some benchmarks, which does have significant value to those readers comparing things on a general level, I wanted to establish a method of testing and reporting that looks directly at the edge of the item’s capabilities. How far can a motherboard be pushed, overclocking-wise, and what is REALLY happening when the board can go no further?
Basically, I want to find, through specialized equipment, where a board no longer functions reliably and, to the extent possibly, identify the reason for the failure to the component or specific functional level (eg., a PCI bus failing the burst mode transfer test at the chosen FSB might indicate that memory and CPU are functioning properly, but the failure is pointing to the Southbridge functions specifically.)
In order to find the edge of capability, we’ve chosen to use a simple test environment that incorporates the excellent products from Ultra-X, Inc.. The test environment consists of 2 specialized PCI cards from Ultra-X. The first one is the RST Professional(Ram Stress Test) card.
This card, without the presence of an operating system, installs in a PCI slot and, after a successful POST, automatically and exhaustively tests all of the RAM installed. Of course the motherboard must have a CPU, heat sink-fan assembly, RAM, a video card and keyboard. No disk drives are necessary and in our testing, we won’t be using any.
In addition to the RST Pro card, Ultra-X produces the P.H.D. PCI test/exerciser card. This card, similar to the RST card, installs in a motherboard and automatically tests each of the motherboard’s functions that the RST card does not test.
This board also monitors the system’s POST and reports any difficulties via a pair of 7 segment LED’s, similar to those used on many newer motherboards. Over the course of the next few reviews, I’ll attempt to delve further into the workings of the 2 test cards and hopefully all of us can learn something about them in the process.
On to the Iwill P4HT.
As I mentioned earlier, the P4HT is RED!!! The board itself is red, as well as the PCI and AGP slots, IDE RAID and floppy connectors, DIMM slots and even the Pentium 4 heat sink bracket. The IDE connectors (the 2 that are supported by the Intel 82801 Southbridge) are blue, as are the LED’s in the cool, but noisy, fan mounted to the PINK 82845PE Northbridge heat sink.
The layout of the P4HT is a little odd compared to other modern motherboards. I can’t say that the layout is particularly bad, but there are some things that I’ve not seen implemented before.
For example, the board has 5 PCI slots (plenty for most people) but the placement of the IDE RAID connectors is at the bottom of the board where a 6th PCI slot would normally be placed. The cables attached here will certainly get in the way of the front panel headers which are located vertically in the bottom right hand corner of the board.
Another unusual layout issue is that there is a full USB connector (not just header pins) located between, and slightly to the right of, PCI slots 1 and 2. I suppose that this makes connecting internal USB components a little easier, but its height and placement make it difficult to use:
Adding to the odd USB placement, Iwill decided to use header pins for USB 5 and 6 (not unusual), but placed them directly between the AGP slot and the first PCI slot. If you intend to use a large video card (like a Gforce4 Ti or similar), the placement shouldn’t be a problem since the GPU fan and heat sink is far above the tops of the pins.
If, however, you are using a smaller video card (eg., Matrox or Gforce MX), there is only about 8mm (app ¼”) clearance between the top of the pins and the bottom of the video card’s heat sink, thus making the header unusable.
The only other layout issue worth mentioning is that there is a row of 6 large capacitors touching the left edge of the Pentium 4 heat sink bracket. This will pose a problem for those of you whose heat sinks have any degree of overhang from the bracket (like the Zalman flower heat sinks).
The Iwill P4HT employs the Phoenix/Award BIOS. Everything an overclocker might want in the way of tweaking is available in this BIOS; however, the maximum voltages available are 1.85vdc Vcore and 2.9vdc for the Vdimm. Personally, I feel that these maximum voltages are sufficient; however, some of you might feel the need to pump more voltage into your CPU and DIMMs.
FSB settings, like most up to date boards, can be done in 1 MHz increments from 100 to 248 MHz. RAM settings available are CAS 1.5, 2.0 and 2.5, and 5, 6 or 7 for the Active to Precharge precharge setting, along with the standard 2 or 3 for RAS Precharge and RAS to CAS delay.
Memory multipliers are rather limited, with only a 1:1 and a 3:4 setting available, along with an “Auto” setting. The available memory multipliers do not change with the FSB setting – the same 3 options are available regardless of the FSB setting.
The Auto setting, at least in all of the testing that I did, always resulted in the 3:4 multiplier (3:4 means RAM clock = FSB*1.33, eg., if FSB=100, then RAM clock=133). Setting the RAM to operate synchronously with the FSB REQUIRES you to set the multiplier at 1:1.
Iwill, however, doesn’t refer to these settings as “multipliers”, “memory clock” or “ratios” like many other manufacturers. The BIOS setting is called “Memory Frequency for:” and the options are DDR200, DDR266 and auto. This isn’t a “bad” thing, just a little different from other approaches.
- Unlocked Pentium 4 1.7 GHz Willamette CPU
- Retail Pentium 4 2.53 GHz CPU
- 256M Corsair PC3200 CAS2
- Vantec 470 watt Stealth power supply
- Liebert 2.7kVA uninteruptable power source (UPS) with integrated power conditioner
- Matrox G550 AGP video card
- Vantec Aeroflow heat sink fan assembly with 5800 rpm TMD fan
RAM and PCI Stress Testing
My approach to testing using the Ultra-X cards is to first verify proper and consistent operation at default settings (voltages, FSB, CPU multiplier, etc.) of the motherboard. In this case, our Pentium 4 CPU, although unlocked, has an FSB of 100 and a multiplier of 17.
With these settings and the memory set to operate at DDR266, the board passed numerous loops of both the memory and the PCI tests. No errors were encountered in 11 hours of memory testing and 4 hours of PCI testing (The RAM tests and the PCI tests use different Ultra-X cards and must be done separately.)
Overclocking Stability and Analysis
My approach to testing for stability is to try to determine the maximum FSB setting that the board will operate under without any errors reported by the RST or PCI test boards. In order to do this, we want to make sure that the CPU and the RAM are NOT factors in limiting operation. This, however, is impossible.
We can only do our best to create the conditions where these 2 critical items “probably” are not the limiting factors. The unlocked CPU is changed to a very conservative multiplier of 10 – the lowest available. The idea here is that the CPU’s default is 100FSB * 17 multiplier = 1.7 GHz, but 170FSB * 10 multiplier ALSO equals 1.7 GHz.
If the CPU can operate reliably at 1.7 GHz (it shouldn’t care if that figure is arrived at via 100*17 or 170*10), then a failure would indicate that the CPU was NOT the problem. Remember though, that this is just the starting point for testing maximum FSB.
As for the RAM in this setup, the same approach holds true. Default RAM settings mean that the RAM is operated at the settings contained in the SPD on the DIMM via the “Auto” setting in the P4HT’s advanced chipset options menu. In order to do our best to try to get the motherboard to fail from overclocking BEFORE the RAM, we deliberately underclock the RAM by using the synchronous 1:1 ratio (for DDR200 in the case of the P4HT) and use very conservative settings of CAS 2.5/3/3/6.
So, just to recap a bit, our beginning setup for stability testing is:
- FSB=170 (ONLY item to be varied)
- CPU multiplier = 10
- Memory = 1:1 for DDR340
- Vcore, Vdimm, Vagp = default
- PCI/AGP clock=locked at 33/66
With these settings, the board performed flawlessly (with one slight exception discussed below) on both the RAM stress tests as well as the PCI tests. Amazingly, the point at which the board started to have consistent memory errors was with the FSB set to 204 MHz (DDR408). The RST card reported a FEW (maybe 9 bad bits) errors at this speed. When I backed off the FSB to 201 MHz, the board was able to operate for over 12 hours without any reported errors.
So, I can safely say that this board will reliably overclock, using the approach described above, to an FSB setting of 201 MHz with the RAM running synchronously at DDR402. In my opinion, this is pretty impressive, particularly in view of the fact that all voltages are set at their defaults and we are using unspectacular air cooling.
One of the areas that I will stress in the upcoming reviews is not simply the point at which a board will not operate reliably, but to the extent possible, I will try to determine what area of the board is failing.
At the very high FSB setting where this board begins to show errors, the failure mode that I witnessed was, as I said above, very small numbers of memory read errors. Since the RAM has been set so conservatively in this round of testing, we can feel confident that the physical board and/or the Memory Controller Hub circuitry were the areas of weakness (although a 201 FSB can’t really be considered “weak”.)
Since it looks like we’ve reached one of the 2 important failure points (the threshold of FSB where reportable errors begin to happen with the RST tester), we now need to see what the PHD PCI test card has to say about these settings. That is, can the rest of the P4HT operate reliably at these levels?
Upon installing the PHD PCI card and booting up the system, the tests start automatically. Each test is displayed on the screen and the number of the last test passed appears on the board’s 2 character LED display. With the P4HT still set at the settings detailed above (201 FSB, etc.) the FIRST run of the tests passed without error.
The second run of the tests (they loop automatically until you stop them) failed, but in a manner that seemed to indicate that something might be wrong with the tester, and not the motherboard being tested.
The tester reports the number of the last passed test along with a single LED that indicates pass, fail or skipped. There is an option for the tester called the Hummer IP, which is an add-on board that exercises functions that the PHD PCI cannot perform on its own.
In our setup, the Hummer IP isn’t used. Ultra-X’s documentation states that tests that rely on the Hummer IP will be skipped if the board is not present. Very oddly, when running the test suite the SECOND time, the tester consistently hangs during the PCI burst transaction test. The odd aspect about this is the fact that this particular test relies on the not-present Hummer IP card and should be skipped. I am in contact with Ultra-X about this, but unfortunately everyone who can help is at Comdex in Las Vegas this week.
So, seeing that I couldn’t run 2 consecutive passes of the PHD PCI card’s tests, I was forced to reboot the system after each test run. Again, each time the tests were run the first execution passed with no errors reported, but the second run hung.
Unfortunately, I have to reserve judgment about whether many single runs without failures is the same as many continuous runs without errors (One important fact to note is that the odd behavior is exhibited with the PHD PCI regardless of the FSB settings. The SECOND run of tests hangs at the same point in the test suite every time).
- The BIOS PC Health Status screen reports temperatures that cannot be correct (eg., 15C CPU temp in a 22C ambient room). Its ONE fan monitor channel for the CPU fan doesn’t recognize our TMD fan. The voltages reported all seem quite high and require verification with equipment that we will be receiving soon.
- This Iwill P4HT and my Samsung PC3200 CAS3 memory do not get along with each other. Regardless of the RAM, FSB, voltage or other settings, this RAM stick will not work with this motherboard. The board doesn’t POST and the speaker emits a constant loud beep. This RAM works fine in other motherboards.
- Unlike many other Pentium 4 motherboards I’ve worked with, this board seems to have a very small gray area between rock solid operation and not working at all as the FSB is increased. There were very few times when the board POSTed and then failed later.
Typically this board would either POST successfully and completely, then execute and pass all tests, or it wouldn’t POST at all and would need to have the CMOS cleared. As the FSB is increased to its threshold, the system seems not to fail subtly – it just refuses to post.
- The LED fan is a nice esthetic touch, but it’s rather loud and doesn’t match the color scheme of the rest of the RED!!! board.
- The white lettering on the RED!!! board is difficult for my old weary eyes to read-even with my glasses.
- The P4HT comes either with Serial ATA (via the Promise SATA controller) or with simple PATA RAID; the board we received for testing is the latter.
- This board uses Hermei and Sanyo capacitors.
Overall, this motherboard, at least in the limited 3 days of testing so far, seems to be quite a solid and stable board with excellent overclocking potential. Adding lots of PCI cards will be somewhat of a problem due to the location of some of the various connectors, but some forethought about where to put the cards might mitigate the problem.
Over the next few weeks, we’ll be getting some additional test equipment here at Overclockers.com and I’ll be adding to this review. I’m particularly interested in doing oscilloscope analysis of the power circuitry when pushing this, and other motherboards, to their limits.
Please feel free to contact us if you have some suggestions and ideas that you’d like to see in our evolving test suite.