I grabbed a Gigabyte board this time around – the G41MT-ES2L. In many ways, it’s similar to the Asus in that it’s got onboard video, a few weak overclocking options, and is about the same size. The largest difference really (aside from using a different Intel chipset) is that it uses DDR3.

—

The Gigabyte went a fair chunk higher – 343Mhz on the FSB. There seems to be a solid wall at this frequency.

UPDATE: I managed to squeeze out another 5Mhz (348Mhz) after doing a little research although it’s important to note it was NOT STABLE (failed Linpack) – based on posts and reviews I’ve come across it seems the G41 chipset is known for a wall between 340-350. In some cases, by playing with the PCI-E frequency setting you can get a few more Mhz out of the chip. 103-105Mhz PCIe worked for me and allowed me to get those extra 5Mhz. Anything above that gave a blank screen. Note that you can really mess up the system (possibly toast something) by running the PCIe bus out of spec, so I suggest being careful if you go this route – remember, you probably won’t get above 350Mhz FSB, so whether spending the time tweaking is worth it for the 7Mhz or less extra you might get is debatable. Try to keep the PCIe around the default of 100Mhz if you can (110+ is risking it), and save this step for the end when you’re sure you’ve hit the wall.

A few details as to how I got there:

• With the E2140 plugged in, simply bumping up the voltage slightly, 300Mhz was easily attainable.
• Moving up to 333Mhz technically *worked*, but the onboard video freaked out as soon as Windows started – the video kept resetting along with messages about the video recovering from an error.
• Using the BSEL mod, onboard video was stable at the Windows desktop. However, at 343Mhz, I hit a wall. No amount of voltage or tweaking got past it – it simply works perfectly fine at 343Mhz (with only the CPU voltage bumped up slightly), and refuses to boot after that.

A few notes about the BSEL itself:

• You’ll commonly hear that Gigabyte boards don’t work with the BSEL. However, with this motherboard that’s only half true. If you boot with Auto BIOS settings, the board will indeed boot at your selected BSEL – both CPU-Z and the BIOS boot screen will reflect this. However, when you hit DEL to get into the BIOS settings, the settings pages will show/reflect a 200Mhz bootstrap. Multipliers (for RAM, etc) and everything else will look the same in the BIOS settings, but you will in fact be running multipliers for your selected BSEL/bootstrap. As an example, you’ll have to check CPU-Z to see what your memory is *actually* running at, because the BIOS settings will show wrong frequencies and multiplier.
• Using the 200 -> 266 BSEL (800 -> 1066), the board boots. Any time I messed something up with the overclocking settings to where it wouldn’t boot, it would recover with “default” settings (using 266 as the FSB).
• Using the 200 -> 333 BSEL (800 -> 1333), the board boots if voltages are set up in the BIOS beforehand. Any time I messed something up with the overclocking settings to where the board wouldn’t boot, it would *try* to recover, but would just beep really fast over and over. (I had to swap in my other un-modded E2140 or scrape off the conductive material used for the BSEL to get it to boot again). Presumably, the default chip voltage isn’t enough to allow it to boot with the 333 BSEL. Doing the volt mod along with the BSEL mod might help overcome this.
• Using the 200 -> 400 BSEL (800 -> 1600), the board didn’t boot at all, even with a voltmod to 1.45V. Again, it wouldn’t recover at all, and I needed to either swap in my other chip or scrape off the BSEL mod so that it would boot at default values.

In any case, the BSEL does work to change the bootstrap. Proof is in the way that the onboard video wouldn’t die at an FSB of 333Mhz if there was a BSEL mod (whether 266 or 333), but would constantly reset at the default 200.

However, the BSEL won’t unlock any memory multipliers in the BIOS (although it will *change* the actual memory multiplier used which you can see in CPU-Z).

—

So is it worth using the BSEL?

I’d suggest sticking with an un-modded chip to start. If you’re using onboard video and it starts acting up, use the BSEL mod (or just plug in a real video card). If your RAM starts limiting you, use the BSEL to get a different multiplier.

However, keep in mind that you may very well hit an FSB wall. It’s very obvious when you do – things work perfectly fine (OCCT/Prime95 stable) and 1Mhz later the board won’t even boot. No amount of BSEL modding is likely to help you when you hit that point.

The best suggestion if overclocking with this board is probably to find a chip with a low FSB and a high multiplier so that you’re more likely to hit the overclocking limit of the chip well before you hit a wall (for reference, the E2140 has a small 8.0  multiplier and a base of 200Mhz). This goes double if you’re planning on using the onboard video. Choose the right chip, and you won’t have to play with any BSEL modding.

If you’ve already got a chip that isn’t optimal, BSEL modding can certainly help, although memory and onboard video are the only things it’s likely to help you with – if your problem lies elsewhere, your mileage with BSEL may vary.

High-End Air -VS- Low-End Water (yes that's my ugly drawing of a fan and a glass of water)

Sorry about the pics – CoolIt doesn’t have images of the Domino on their site anymore, and I was too lazy to take a pic. There are a zillion pics on the web, but I didn’t want to rip one off. Hence, the barely-representative-but-you-get-the-gist-of-it drawing.

Anyway…. little blurbs, then benchmark numbers at the end.

—

NOCTUA NH-U12P

Noctua’s coolers are generally considered to be fairly high end. While the U12P isn’t the latest model, it’s still quite good, and I used it for quite a while in the overclocked AMD X6 system.

Note that the Antec 300 case it was in has excellent airflow. It was mounted pushing air up (1 fan), with the case’s 14 cm fan just above pulling that air up and out – the ideal setup for this case. When the system was under load, you could *feel* the warm air being pulled out the top.

Like I said, this cooler’s pretty good. To give you an idea, it’s “new” home (an i3), it’s dropping load temps by 20 degrees celcius… from an alarming 92 down to 72 – I’m sure it would do even better if the “new” home had better airflow too, but 20 degrees is nothing to scoff at – it’s impressive.

—

COOLIT Domino A.L.C.

This is what’s known as “low end” watercooling – essentially, the all-in-one package priced (and meant to compete with) high-end air. This falls along the same line as the popular Corsair H50 / H70 water systems. In fact, when I hopped into the store, the choice was between this and the H50.

In comparison to the H50, most of what I’ve read seems to indicate that the Domino beats the H50 by a couple degrees. Every test system/set-up is different though, so that can be taken with a grain of salt. In any case, it should be fair to say that they’re both similar in terms of the performance you can expect. In any case, the Domino was a few dollars cheaper, and had an LCD which displays (amongst a couple other things) the coolant temp.

I should note that if you’re looking at the Domino vs the H50, keep in mind that Corsair’s a very popular and solid name and in terms of trusting a brand not to leak all over your system, Corsair would probably be the more popular pick. The Corsair also allows for 2 fans (in a push/pull configuration), whereas the Domino only has room for 1 fan (due to the LCD/pump section taking up the space required to add a 2nd). Corsair defaults to sucking air into the case, whereas Cool-It defaults to pushing air out. Both orientations can be changed.

—

TEMPERATURE RESULTS

You’ll see the CoolIt Domino comes out ahead. HWMonitor was used for the number collection (the numbers correspond to what both RealTemp and AMD OverDrive report). To get the max values, OCCT was run for 15 minutes, then the “LinPack” module was run for another 15 minutes. Then I waited to see what the lowest idle temp was and took the min/max numbers.

Note that AMD Overdrive refers to these values as follows:

CPU = Core Temperature
TMPIN0 = CPU Temperature
TMPIN1 = Motherboard Temperature

—

RESULTS

Depending on whether you believe CPU or TMPIN0 to be the most accurate representation of the processor’s temperature, the Cool-It was better by 2 degrees at idle, and 4-6 degrees at load.

You’ll notice that TMPIN1 also saw improvement by a flat 3 degrees (whether at idle or at load). The reason here is simply a change in the case’s airflow dynamics caused by using a different cooler which modifies the airflow within the case.

An important thing to note is that both coolers used their included paste (NH1 on the Noctua and the default thermal layer on the Domino), and had the fan speeds set on high. Overall case airflow is very high (case is quite loud), so if you test these 2 in a cramped case you’ll probably see different numbers.

-

Note that the CoolIt’s “high” speed is very loud. I mean *very*. I mean the case is already loud and this thing bumps up the volume an incredible amount. I tested the thing at medium and slow settings (at which point I can’t hear it over the other fans), and these were the results:
Medium: CPU=43 – TMPIN1=57 – TMPIN2=39
Low: CPU=45 – TMPIN1=58 – TMPIN2=44

At the medium fan speed, both the CPU and TMPIN1 temperature are exactly the same as the Noctua, although the motherboard temp is even lower than before (again, varied airflow dynamics in the case).

-

CONCLUSION

The Cool-It Domino A.L.C. does give an improvement, but it’s not by much, and it’s at the expense of incredibly loud fan speeds. At medium speed, it performs almost exactly the same as the Noctua.

—

So if you find them at the same price, which should you buy?

It really depends. They’re both good performers.

First, ask yourself if you’re ready to deal with watercooling. Remember if the Noctua’s fans die, it’s an easy replacement. If the pump in the Domino dies, you’re looking at an RMA if you’re within the warranty period, and if you’re out of the warranty period, you’ll have to junk it.

In addition, water cooling always has the potential for leaks, which can devastate a computer. I had a ThermalTake water cooling kit years ago, a seal in the pump leaked, and it burned/melted/fried internally dripping plastic in the case, leaving a stench for days, and also killing a motherboard fan header in the process (fortunately nothing else). Remember, with air cooling the worst thing that can happen is generally a fan dying (or more rarely, an improperly mounted heatsink falling and smashing something).

Next, look at your case’s airflow. Would the Noctua be blowing heated CPU air into your PSU? Might it block or reduce airflow over other motherboard components? The problem with air coolers is that they cool the CPU, but that warm air still has to get out of the case somewhere. With the Cool-It, the radiator/fan get mounted right in the exhaust port, so you’re moving all that warm air from the CPU directly outside the case, and the waterblock’s small enough that it’s not restricting airflow around components situated near the CPU.

If your case temperatures are high, the Cool-It is undoubtedly going to be your best bet. If your case temps are already low or if you don’t think the risk of water-cooling is worth it, then the Noctua’s a great choice too.

-

I’d have no problem recommending either of these coolers. They’re both so similar in performance that the “best one” is going to depend mainly on your current case and airflow situation.

I have to say, it’s probably the best printer I’ve ever owned.

I’ll begin by mentioning that I’m not going to focus on speed, color representation, or any of the other stuff that most printer reviews focus on. I’m a pretty regular guy when it comes to my printing needs. I don’t hundreds of pages per day. I’m not an artist. That said, I want a printer that works well and quickly the times I use it, doesn’t cause nightmares when shared over the network, and also isn’t going to cost me an arm and a leg in toner over the course of it’s life.

The Brother HL-3070CW fit the bill perfectly.

I should also mention that I’m the type to pay 3X more for a printer if toner/ink is cheap (or if it can be refilled through generic eBay toner/ink). Naturally, this means I’ve never bought a Lexmark and never will. This also means I used to buy Canon but never do anymore. You want to chip your printers? Have fun. I won’t buy from you. Ever. Brother’s one of the few manufacturer’s left that I’m willing to buy from.

—

In any case, with the Brother a seemingly good fit on paper, I’ll keep this short and give a few pros/cons:

PROS

• Works great with the Mac. Whether plugged in directly to the computer through USB, through Ethernet, or via the Apple Time Capsule, Mac OS X (Snow Leopard) detected it and automatically downloaded drivers through Apple’s Software Update.
• Can be configured through a web interface if you use it as a Network printer. Loads of options here too.
• No draconian anti-refill measures put in. If you’re the type to buy refill toner from eBay, it’s a mechanical process to refill/reset the cartridge (no melting or chip required – just a couple screwdrivers), and it’s much easier/simpler than some other brands where you have to melt/glue/tape/rechip/etc.
• Reasonably priced “Brother” toner. The cartridges seem to hit around $50 bucks a piece on Amazon, which is lower than most toner cartridges from other manufacturers. Since I buy eBay refill kit stuff this doesn’t really apply to me, but for those who don’t want the hassle of refilling their own cartridges, paying a reasonable price for the real thing is always good.
• It’s nearly impossible for dust / pet fur / etc to make it’s way inside the printer. Paper tray’s hidden (and horizontal), and the output is horizontal also. Examine the picture – it’s impossible for junk to get in there and mess with the rollers over time.
• Wireless. While I have it shared through the Time Capsule (USB), you can set it up wirelessly. Or, you can plug it in to a network cable. Either works.
• Silent when not printing. After 5 minutes (the default time, you can change this in it’s settings), it goes to “sleep”. Lights go out, fans shut off, and it uses under 10W. It doesn’t randomly start up (like some other lasers I’ve had). Unless you look at the display you’ll think it’s off. I like it.
• Starts up rather quickly from sleep. When I print from a computer in the other room, by the time I get to the printer, the pages are usually already coming out. This is in contrast to just about every other laser (and inkjet) I’ve owned where it seems to take a couple minutes of prep before it actually starts printing.
• Low sleep/idle power consumption – it uses 7 or 8 watts as measured from the wall when in “sleep” mode. That beats the other laser’s we’ve had. While there’s always room for improvement (I’d love to see a printer hit < 1 watt at idle one of these days, matching computers in S3 mode), it’s still very reasonable.

CONS

• Network setup – if for whatever reason you choose to do it entirely through the printer’s built-in interface (rather than the Brother software), it’s a nightmare. You have to be very knowledgable when it comes to DHCP, IP addresses, subnet masks, etc. You can’t do it entirely through the menu buttons on the printer either – you’ll use that for part of the setup (wired) and then log in to the web interface (to set up the wireless part).

  Really, you’re best just to use the built-in software to set it up on the network. The printer’s interface (and web interface) is sub-par when it comes to a wireless network setup. There was just enough effort put in to make it technically possible to do it that way – Brother really expects you to use their software to set it up on the network instead.

• Large – You’ll start to get an indication as to how big this thing is when you see the box it comes in. In terms of size, having a completely horizontal design for paper output really doesn’t give them options in terms of width/length. It’s a tradeoff. You won’t have dust getting in there, and the paper will come out and rest nice and flat, but it’s going to take up a chunk of space on your desk. Personally, I think it’s a good tradeoff, but if you’re limited for space, you may have to make sacrifices and go with a vertical-style printer. However, the printer’s not very high, so if you want to put it in a shelving unit of sorts, it may be perfect.
• Heavy – This won’t matter for most people. Most people will need a friend to help carry the thing to and from the car when they buy it, but once it’s at home and installed, it’ll sit there. However, if for some reason you need to move the printer often, this could be a problem for you. It’s bloody heavy and is easily the heaviest printer I’ve ever owned.

Really, it’s a solid printer, and I don’t have any regrets.

While other printers may beat the HL-3070CW in terms of initial price, the Brother comes out very well when you look at the overall long-term cost including consumables – particularly if you refill the toner yourself. Brother’s well known for their quality and service too, so you really can’t go wrong here.

If the built-in interface for setting up the network was a little more versatile/polished, and if power draw when idle dropped to 1-2W, I’d give this thing a 10/10. However, those are relatively minor things to nitpick about. It’s worth it’s weight at regular price, and at the sale price of under $200, it’s a steal.

In the past, I’ve used an Antec Sonata III (review here), and I had the Sonata Piano finish before that. Overall, I’ve been pretty pleased with Antec cases.

—

I skipped out on the Sonata this time around and went with the 300, because massive cooling potential was the goal this time around. While the Sonata’s do cool very well, I was looking for something to push more airflow.

I’ve taken some pictures with the Antec 300 pulled apart, with a few comments attached to each to give you an idea as to my thoughts on this case.

Here you see can see the front panel when removed. It’s fairly easy to get off, although you’ll only need to remove it when adding front case fans (a couple 12cm fans are supported), or when removing the screen filter for cleaning. To remove it, you pop off the side panel, and then push in the black clips that hold it on – basically the same way you’d remove a front panel on a standard case. There aren’t any wires or anything attached (unlike cases that have the buttons/leds integrated with the panel), meaning that once it’s pulled you can bring it to the counter without having it tied to the case.

The filter is a bit of a pain to remove. Looking at the image, there are 4 clips along the left side, and unless you’re fortunate enough to have 4 hands, you have to pull up on the screen’s frame while pushing back each clip. Once the screen’s removed, you can clean it easily enough, although putting it back in is actually tougher than pulling it out – it doesn’t snap in very quickly/easily, and really requires a lot of force, leading to concern that either the frame might crack or the clips may break.

This is in contrast to the Sonata’s filter, where you simply tip the case on it’s side to access the bottom, squeeze 2 clips, and slide the thing right out, and back in again.

Fortunately, you probably won’t be pulling the filter more than once a year or so, but it’s still something that should be easier. Thinking you might break something is never good.

Moving on to the 12cm fan cages, they’re pretty awesome.

You can just make them out, but at the bottom (of the picture), you’ll see 4 thumbscrews – 2 for each cage. Unscrew them, swivel the cage up, and it comes right out. You can then install the fan into the cage without having to work around the constraints of the case. Once you’ve got the fan screwed into the cage, the cage just latches on, swivels back down, and you reattach the thumb screws.

It’s really that easy, and even though it’s something you only tend to do *once* (unless a fan starts buzzing), it’s a very well-thought-out design. A+ here.

Now we start delving into the case. You’ll notice I’ve already got the motherboard + CPU cooler installed.

The Antec 300 has both a rear and a top exhaust port. The rear is a standard 12cm fan (comes with the case), and the top is a 14cm fan (also comes with the case). Both fans have speed switches with Low/Medium/High so that you can choose the balance you want between noise and airflow.

One thing to note is that due to the position of the 14cm top fan (and the size), you’re actually going to get some airflow over the MOSFETS on the motherboard (just to the left of the CPU cooler). This is a pretty big plus – there often isn’t much airflow over these components unless you’ve got certain CPU coolers that pass airflow over them. Because the rear fan is always elevated above the motherboard level (necessary due to all the rear ports), it’s never able to pass airflow directly over the motherboard. The large top exhaust fan remedies this issue. Depending on the orientation of your CPU cooler, the top fan can also aid in keeping your CPU heatsink cool (and as you can see, this applies to the Noctua cooler I’m using).

I’ll mention briefly that top/rear fans should almost always be exhaust fans, except in extremely rare circumstances. In the default orientation, these are.

Here’s a larger shot with a little more. I probably should have spun the orientation, so for those who may not be completely familiar with case layouts, “up” is the left side, and “down” is the right side.

You’ll see the power supply mounts at the bottom of the case (bottom-right in the pic). There’s a bit of clearance between the bottom of the PSU and the case, which is good because an awful lot of PSU’s have bottom-side intakes, and blocking that off would definitely result in overheated PSU’s. Well designed here.

In other cases that mount the PSU at the top, I’ve often used/considered the PSU to be simply another exhaust fan, and initially I was a little leery on going with a bottom-mount design. However, I can certainly see the benefits. With the PSU at the bottom, it’s not sucking in any warm air, which means if you really load the PSU, it won’t have to deal with the additional case heat being pulled through it. Remember, in a top-mount design, if you have a CPU that’s pumping out piles of heat, the PSU is always sucking in some of that heat – and that’s before adding it’s own heat.

This really is a better design. The PSU gets it’s own cool air, and it’s still spitting it’s heat out the back, away from the other components.

Next, you’ll notice the drive bays are the standard screw-it-in style. I’m not a big fan of the screw-it-in style (I prefer methods to mount the drives quickly). However, with the motherboard mounted high up, there’s ample room for a number of hard drives, without worrying about them hitting motherboard components when you’re trying to put them in.

Note that the front grille (and 2 12cm fans should you choose to install them) is directly in front of the 3.5″ drive bay. Hard drives tend to cook, so having the intake start here is definitely the best solution, and if you do add intake fans your hard drives should stay quite cool indeed.

Finally, a shot with the side panel (almost) on. It’s not a flaw, I was just working in the case still and didn’t snap it down since I’ve usually got wires dangling out.

There’s room for a 12mm fan on the side panel, and you’ll see I’ve used the 2nd Noctua fan here as an intake.

This is one area I’m not thrilled about due to the lack of a filter. You’ve got a filter on the front, but the side panel’s going to be blowing in dusty/linty air. Filtering half the air doesn’t make a whole lot of sense, but unless you construct your own filter of sorts for the side, that’s how it’s going to be.

You could just *tape* over the side, but it’s kinda ugly. You’re also missing out on airflow over the video card and lower chipset area (possibly catching the southbridge) if you tape it off.

The other option is to use it as an exhaust port, but that makes about 0 sense. If you use it as an exhaust, you’ll have 4 exhaust fans (including the PSU), and only a max of 2 intakes. Unless you went to huge lengths to balance the airflow (high speed intakes, low speed exhausts), you’d probably end up with a lot of unfiltered air being pulled through all the seams/nooks/crannies of the case (which ends up filling those seams/nooks/crannies with dust). Even if you balanced the airflow though, it would still be useless. You’d be exhausting air that hasn’t cooled anything but the hard drives, and that warmish hard drive air would still be passed through the rest of the case anyway.

It’s an unfiltered intake. That’s about all their is to it. You can tape it off, lower the fan speed to reduce the air pulled from it, create a filter, or… most likely… just deal with it.

—

Final Thoughts

Really, for a low price case (well under $100), the Antec 300′s a pretty solid purchase. You do get a filter, a couple free speed-adjustable fans (one of which is a whopping 140 mm), and a very decent design in terms of airflow. The biggest shortcoming is the unfiltered side port, and that’s really quite minor all things considered. It would be nice if the front filter was more easily removed/re-installed, but for something I’ll do once every 1-2 years, I can’t really complain – I’m just glad it’s there.

This PSU is going on a field trip! To the dump!

Years ago, we learned not to buy no-name PSU’s. They’d die during the summer.

Since then, I’ve read reviews, done research, and attempted to go with good solid name brands. Things have gone much better since then.

Well… until this thing came along.

It was bought years ago to run a system. The system needed about 300-400W worth of power. The 600SXS was bought because it was a 600 watt power supply and… well extra headroom is good!

Surprisingly, one day while the system was running, the thing just cut out. After waiting a short time, I started up the computer again, and not long after it cut out again. Fortunately, I had my 400W Fortron Sparkle PSU sitting around, which worked as a replacement. That’s right. FSP. Fortron Sparkle. The ugly low-cost power supplies that happen to be decent. They beat this thing with their 400W model.

Much later on, I was putting together a system for a writeup here, and tried to give the OCZ another chance. The system used under 200W at load (measured from the wall), plus an ATI HD 4850 video card (which is probably in the neighborhood of 150-200W. The power supply died out again, but this time for good. It wasn’t running games, or even stress testing at the time, it was booting Windows. So maybe 250W power draw coming from the system total.

I’ll note that all my Antec PSU’s and Sparkle/Fortron PSU’s have handled loads well beyond what this thing was ever subjected too, even when they’ve been rated lower (400W/500W levels)

—

Pros:

• It was pretty quiet when it worked.
• It shut off safely when cutting out (some no name PSUs will just keep going and start blowing caps until it shorts out or fries your motherboard). This is actually really important – while I’m unhappy the thing died, I’m very pleased it shut off rather than destroying my expensive hardware so I’ll put some gold stars here. —> *****
• I was able to salvage the 12cm fan before throwing it in the trash.

Cons:

• Only lasted a couple months.
• Didn’t handle anywhere close to the max load.

Conclusion:

This thing is not worth buying. To be fair, our house gets pretty hot in the summer, although all the other PSU’s have handled the heat just fine. Fortunately, NCIX doesn’t sell it anymore. There’s a new “version 2″ they now carry – I really hope it’s better than this one was (for the sake of OCZ customers), although I’m not willing to find out for myself.

This time I’m looking at a more complex scene (or at least… a more complex render of it). I took the same scene, chose a section of frames to render that tended to be slowest, and cranked up a few settings in the mental ray options. I bumped up the resolution and added motion blur. Now instead of seconds, I was talking 2-3 minutes to render each frame.

The goal was to reduce the effect of network overhead on the results, with more time spent rendering, and a much smaller percent being taken up by Mental Ray Satellite’s network distribution.

The results were… interesting.

The same 36 frames were rendered for each run, and I used the results from 35 of them (using the 1st as a time stamp).

Note: ignore the last 4 columns - it's just data that I'd added to the chart to look for correlation/scaling. Focus on the 4th data column ("improvement as % of solo render") to see the benefit/decrease of additional machines. I apologize for not simply highlighting that column in the image.

A few notes as they relate to the above chart:

• The “Master” was the AMD 6-core, clocked at 3.5 Ghz.
• For the “1-slave” test, the o/c’ed dual-core 2.32 Ghz machine from part 1 was used as the slave.
• For the “2-slave” test, the i3′s from part 1 (3.2 and 3.06 Ghz) were used as the slaves. Both are similar performance-wise.
• For the “3-slave” test, the 3 machines from above were used as the slaves.
• For the “4-slave” test, a 2.26 Ghz Core2Duo (also from part 1 of the writeup) was added to the above. This additional machine is similar in performance to the machine from the “1-slave” test.

Unfortunately, the test results are riddled with inconsistencies, and you MUST keep that in mind when comparing for your own purposes. The biggest issue is that every machine was different. That means you can’t really compare the results between “1 slave” and “4 slaves” in terms of how much value you’ll get in going from 1-to-4 machines.

Really, you’ve got:

1. An AMD 6-core machine at the highest clock. No hyperthreading, and an AMD architecture.
2. Two intels on the i3 architecture. Both have hyperthreading. Similar in performance to each other.
3. Two intels on the Core/Core2 architecture. No hyperthreading. Similar in performance to each other.

You’d have to be insane to take the “core improvement” and “Ghz improvement” to heart. They’re there simply for informational purposes, and to look for correlation between the work done and the cores/Ghz provided. It’s in no way meant as a set-in-stone-for-every-x%-Ghz-improvement-you-will-get-Y%-more-frames-done thing.

—

Looking at the results (even though we know they’re not perfect)

The first thing to note is that we *still* hit a wall in terms of improvement where things simply get worse when enough machines are added. 3 slaves gave no improvement, and 4 was slightly worse. This could be due to the Core Duos being slower than both the master and the i3′s. However, unlike Part I (where the drop was so bad that it was the same speed as a solo render), we still exceeded the speed of the Master rendering solo this time around, so while it wasn’t helpful, it at least didn’t destroy the efficiency of the others.

One interesting thing to note is how well the i3′s look to have scaled on paper (the 2-slave test). You had a Ghz/Core improvement of 160%/167%, and in terms of the render speed, you were looking at 156%. Really, not that bad. Hyperthreading may have thrown this off a bit (and made things look better) mind you. However, when using a single machine (doing the simple render), the AMD was roughly twice as fast as a single i3. This isn’t mentioned in the data, so you’ll have to take my word for it. In any case, based on that tidbit, you would actually expect that the 2-slave test using both i3′s would have brought a number closer to 200%. It didn’t quite make that.

Further tests..?

While I’m done testing for the time being (it’s time consuming and I’ve used up all the machines available to me), it would certainly be interesting to see what a few other tests show. Tests I would like to run if I had the time/resources are as follows:

1. Testing more than 4 machines, rendering scenes that take a very large base time to render (30 minutes per frame minimum).
2. All machines (master and slaves) the same spec, testing scaling from 0-slaves all the way up to a large number of slaves. This would help to determine the exact scaling in an ideal environment, and also indicate whether some of the slaves being slower contributed largely to the drop seen during my 4-slave tests (both here and in part 1).
3. Reducing the threads on the master to leave 1 core free. Goal would be to see if the 1 open core would do the network distribution, and if so whether letting it have it’s own core would result in an overall improvement.
4. Turning off the option on the master to render on the local machine (only render on the network machines). Goal similar to the above. Trying this on both a fast and a slow master to see what (if any) effect there is, and whether it’s optimal or not to have a master dedicated soley to distributing the render.
5. Disabling/enabling hyperthreading on all machines (in a mixed environment similar to the one I’ve tested with) to see if there’s an effect/impact on the network render.

—

Final Thoughts:

In any case, the benefit in benchmarking before sending out a full-scene satellite render still applies. However, it doesn’t appear to be as critical for longer / more complex scenes – even though utilizing all the machines at the same time wasn’t optimal in my case, it didn’t hamper things as badly as it did for the simple scenes in Part 1.

I’m even more curious than ever as to why 3 and 4 slaves gave a poor result. All the machines *were* working, and the effect of network overhead should have been largely reduced.

ASUS M4A785-M installed with AMD Phenom II X6 1055T - Yes, as you'll see, all those fans are necessary!

Preliminary Warning:

Overclocking has the potential to damage/destroy components. Overclock at your own risk. Just because my settings didn’t blow up my machine, doesn’t mean they won’t blow up yours. The writeup below is subject to error and inaccuracy. If your machine dies, your house burns down, or you inadvertantly cause a chain reaction of events resulting in a nuclear power plant exploding due to something you read here, I disclaim all responsibility.

—

With that out of the way… the M4A785-M isn’t the newest of boards, and pairing it with one of the best AMD processors might seem a little puzzling. Here’s why it was chosen anyway…

• It’s cheap (under $100).
• It supports DDR2 RAM.

The fact of the matter is, I had perfectly good DDR2 RAM that wasn’t being used, and one of the newer AM3 boards just didn’t make sense. Therefore, I needed an AM2/AM2+ motherboard. Even though the X6 is an AM3 processor, it’s backwards compatible with AM2/AM2+ motherboards, proving the board supports it physically as well as through a BIOS update.

—

Pre-build concerns:

• The M4A785-M is a budget board. Budget boards and high-end processors often *barely* manage. Adding overclocking to the mix has the potential to spell trouble (heck, budget boards often have difficulty dealing with overclocked low-end chips).
• While it supports the X6 through a BIOS update, this motherboard obviously wasn’t originally designed with the 6-core in mind.
• It supports processors with a maximum of 125W (which most of the current X6′s are). Again, adding overclocking potentially pushes the limit. Note that there are other boards that support 140W processors (the ASUS M3A78-EM being an example of an older DDR2 motherboard that supports the 140W processors). It would be reasonable to assume that a 140W-supporting variant would be better suited to the task.
• 4+1 phase power design. I suppose it could be worse (Gigabyte has a DDR2 motherboard supporting the X6 with only 3+1), but there are AM3  boards out there with 8+2 phase.
• El-cheapo heatsinks. The northbridge heatsink is actually sized very well. However, the southbridge heatsink is as tiny as possible, and there’s no heatsink on the MOSFETs.

Despite these downsides, motherboard options in the DDR2 realm are slim, and this was the best motherboard locally available.

However, the reasons above are largely why the overclock was done at STOCK voltages. Overclocking adds a bit of heat and power consumption. Overvolting increases the heat/power-consumption substantially and I’d be begging for trouble doing it (I wouldn’t expect the board to last more than a year assuming it survived the overvolting process on this processor to begin with).

Hence, stock voltage results only.

—

Board setup

Before throwing the motherboard in the case, I pulled the northbridge & southbridge heatsinks, and scraped off the old thermal interface junk. It’s silver stuff on the northbridge (similar to the AMD heatsink thermal interface pad), and the gross pink thermal pad on the southbridge. Once those were off, I replaced it with an extremely tiny amount of Arctic Silver.

If you replace the thermal pad with thermal paste, I’d recommend using a NON-CONDUCTIVE paste. The dies on the north/southbridge are so tiny that it would be easy to use too much, and if you conductive stuff you might kill the board.

I also used Arctic Silver on the stock AMD CPU heatsink. The included heatsink is actually pretty decent (heatpipes and everything), but the default thermal pad was too thick for my liking.

The motherboard’s a standard ATX size – if you’re using a medium-sized case, you’ll probably have to yank out your hard drives while you install it.

In terms of layout, the board’s pretty good. The one exception is that the 24-pin power connector hugs the IDE connector – this won’t matter if you’re using SATA, but if you’ve got an IDE drive, getting the cable in will be a little tight.

—

Initial set-up

First, the board needed a BIOS update to support the processor. It detected it as an “Unknown Processor” and gave an error message about a CPU not being installed, although it still let me into the BIOS where I could change options and pop into the ASUS Bios Updater.

The BIOS included was version 702. The current X6′s require at least version 906. I used another computer to download the new ROM onto a USB memory stick, popped it in the M4A785-M, and let it flash. After a restart, things were looking good.

—

I did some quick testing at stock setttings. Note that I had close to the bare minimum attached to this thing, aside from a pile of fans. 1 hard drive, 1 dvd-rw drive, 2 sticks of DDR2-800 RAM, and that’s it. I was using the onboard video and had nothing else but a keyboard/mouse plugged in.

At idle (Windows 7), the system was using 84 watts at idle (measured from the power outlet), and 172 watts at load (running Prime 95). An average power supply should be able to handle this thing if you’re using onboard video.

—

Heat

While Prime was running, I touched the heatsinks. The CPU heatsink was relatively cool. The northbridge was cool (although the RAM cooler was giving it some airflow), but the southbridge was extremely hot. After about 5-10 seconds I had to pull my finger off to keep from cooking my skin. Immediately after shutoff, I felt the MOSFETs and they were hot.

Here’s the concern…

Southbridge – the heatsink is just too thin and tiny. Regular case airflow just doesn’t cut it. The thing’s screaming hot at stock, under load, with good case airflow. I’d hate to think what it would feel like in a stuffy (HTPC) case.

MOSFETs – the CPU fan passes some air over this region, but if you have the silent fan control enabled in the BIOS, at low CPU temps you’re not getting airflow.

I wouldn’t dare to overclock under these conditions. Overclock + a hot day could easily mean motherboard death.

My fixes were as follows:

Southbridge – I used zip ties to hang an 80mm fan over the location (which you can see in the picture), just to get some directed airflow. It worked very well.

MOSFETs – I turned off the setting for the silent fan in the BIOS so that it ran at full-speed all the time for maximum airflow in the area. I also used a dremel on the computer case to cut out the “grill” for the rear exhaust fan, so that the rear fan would pull more air over that general region (and increase total case air flow). This had a helpful effect – the MOSFETs still got very warm, but not as hot as before.

It’s pretty clear that the M4A785-M is marginal at best when it comes to cooling and the ability to deal with heat. I have a RAM cooler, rear exhaust fan, and 80mm fan over the southbridge added to the setup just to keep things comfortable temperature-wise at stock. I would have been very hesitant to risk overclocking without having dealt with the southbridge and MOSFET temps.

What a terrible southbridge heatsink! I'm surprised it doesn't glow red. You'll see part of the orange fan I hung with zip ties in the bottom left. It was either that, or rig something up so that the southbridge would keep my coffee warm.

Those MOSFETs are in dire need of heatsinks. Even with the CPU fan at full speed (blowing air out the sides), the exhaust fan to the left, and the PSU intake fan above, these things stay very warm. This in itself is a huge reason you don't want to crank up the voltage of the CPU. These little guys probably wouldn't be able to take the heat.

—

Overclocking (finally!)

Now we get to the good stuff.

For all it’s flaws, the M4A785-M has a very capable BIOS for overclocking. My last post was about a terrible Intel-based ASUS. This one’s so nice it feels like the old (awesome) ASUS is back in town again. There are some flaws in the way things are worded in the BIOS, but the options are there and are plentiful.

For those new to the AMD side of things (but not new to overclocking processors), here’s a quick crash course on this motherboard….

CPU/HT reference clock – this is the base frequency that all your other stuff is multiplied against. It’s like the Intel FSB, but better (because more things are detached by way of adjustable multiplier).

Processor Frequency Multiplier – you already know what this does. This multiplied by the reference clock = CPU speed.

CPB Control – aka “Core Performance Boost”. This is the “Turbo” function to boost the multiplier of cores when they’re not all in use (single-threaded tasks). You can adjust the multiplier used for this if desired. I disabled it for overclocking because of a couple issues I’ll mention, as well as the difficulty involved in testing each core individually when turbo’d.

CPU/NB Frequency – This multiplier determines your Northbridge frequency (again, multiplied by your reference clock). The default works out to be 2000Mhz. It’s possible to get a speed improvement by bringing this up, although you’re going to hit a wall without increasing the voltage.

HT Link Speed - This is the “HyperTransport” speed. The default is 2000Mhz. For some reason they list a frequency instead of a multiplier. Note that overclocking this generally has almost no effect on your total system speed, and can even result in a DECREASE in speed. Overclocking it can also cause instability. Therefore, try to keep this thing around the stock frequency of 2000Mhz. The frequency listed is what it would be at a STOCK reference clock of 200Mhz. Therefore, when you increase the reference clock, you must DECREASE this setting to keep the actual result near 2000Mhz.

Memory Clock Mode / Memclock value – This is poorly written just like the HT Link Speed. The frequency shown here (when set to manual) is the frequency the RAM will run at when running the STOCK reference clock of 200Mhz. The math you have to do here is nothing short of a nightmare. When overclocking, it’s generally easiest to set this to 200Mhz (which is actually a 1:1 ratio, and would be DDR2-400 speeds at 200Mhz). Once you’ve finished with your overclocking, start increasing this setting (which will change ratios) until you get near your desired RAM speed.

Advanced Clock Calibration – Apparently this helped the original Phenom processors obtain higher stable overclocks utilizing the southbridge somehow. I’ve read mixed things on this, but from the sounds of it, this probably doesn’t benefit Phenom II’s a whole lot, and could cause instability. I have it disabled, but you could try both and see which nets you better results.

There’s plenty more I could go into, but this should be enough for now. In any case, the settings I ended up with were as follows:

CPU/HT Reference Clock (Mhz): 250
PCIE OverClocking: Manual (100)
Processor Frequency Multiplier: x 14.0
CPB Control: Disabled
CPU/NB Frequency: 8.00x (works out to 2000Mhz – later increased a notch)
HT Link Speed: 1600Mhz (works out to 2000Mhz)
Memclock Value: 333Mhz (works out to DDR2-832 if I remember right)
C1E Support: Enabled
Advanced Clock Calibration: Disabled

This resulted in a 3.5Ghz machine – a 700Mhz improvement over the stock 2.8Ghz. However, turbo is off (which by default gives 3.3Ghz on single-threaded tasks), so this isn’t a huge improvement for single-threaded stuff. However, everything was left at stock voltage.

Incidentally, I tried a base clock of 260, but it failed Prime95. It’s possible I could have squeezed a little more out of this by trying 255/etc, but since 250 allows me to maintain the stock 2000Mhz HT (at the poorly named “1600Mhz” setting), I left it there.

Most of the other stuff was left at default values. Note that I used C1E to help the processor cool down periodically for the types of tasks I do – you may want to disable it to try pushing higher clocks.

Important note! It’s been said in numerous places that your NB frequency MUST be equal to or higher than your HT frequency (never lower). Be careful of your settings when changing things around.

Finally, use the AMD Overdrive Utility and CPU-Z to see what each BIOS change actually results in setting-wise. Focus on the CPU frequency, NorthBridge frequency, HT Frequency, and RAM frequency. Make changes a little at a time – because the ASUS bios is a little poor when it comes to being clear about what each setting does/changes (listing frequencies when they should list multipliers for example), tiny changes erring on the side of caution is the best way to go.

—

Oddities Abound!

I have never seen oddities like the ones I experienced here.

First, Memtest86+. Using the latest version I had (4.10), it never detected the actual RAM speed. It always used whatever the BIOS (incorrectly) reported based on the setting chosen. So when overclocked, it would say DDR2 667, even though the RAM was running at DDR2 832 when overclocked.

Next, the AMD Overdrive Utility, and CPU-Z. Initially, I had CPB (the “turbo boost”) setting enabled in the BIOS. However, no matter what frequencies I chose, AMD Overdrive kept claiming that I was running at stock (2.8Ghz), while CPU-Z showed my chosen frequency (3.5Ghz for example). They matched up everywhere else. Since I didn’t want to benchmark to see which was telling the truth (and didn’t want turbo boost limiting my overclock), I just disabled the CPB setting in the BIOS. They also mismatched when I’d manually choose a Processor Frequency Multiplier above 14.0 (basically if I chose the “turbo boost” multipliers), which is why I stuck with the 14.0 and just moved up the base clock.

Here I'm keeping an eye on the temperatures reported in AMD Overdrive while overclocked to 3.5Ghz and running OCCT. Yes, it's an ugly CRT monitor that you're seeing.

—

Final results:

At the settings listed above, I’m at 3.5Ghz. Prime is stable.

For the CPU temp, I’m looking at 19 degrees at idle (it’s cold in the house right now), and 42 degrees at load.

System power consumption (using onboard graphics) measured from the wall is 76 watts at idle, 202 watts at load.

—

Final thoughts

All in all, the ASUS M4A785-M is a pretty decent motherboard in terms of capabilities. My overclocked results are in line with what most people seem to be able to achieve at stock voltages on the default air cooler. I’m sure that if I dared to up the voltage I could hit close to 4Ghz, but there’s no telling if the motherboard would actually survive for very long.

What I liked:

• Excellent overclocking options and a multitude of settings.
• The motherboard actually POSTed with the X6 before the BIOS update, allowing me to update the BIOS (some old BIOS’s used to require you to use a *recognized* CPU to even POST).
• Easy, stable overclock.
• Allowed me to use my DDR2 RAM, and you can set a 1:1 multiplier through the “200Mhz” option, which means even if you had ultra-slow DDR2 533 RAM, your memory would be able to run within spec at base clocks of up to ~266Mhz.
• Quite cheap. At under $100, it easily out-does every single Intel motherboard I’ve ever bought for under $100 in terms of options, features, and overclockability.

What I disliked:

• Southbridge cooling is marginal. No heatsink on the MOSFETs.
• Only 4+1 phase power design makes me leery.
• Some BIOS settings named incorrectly (showing HT and RAM as frequencies, even though they’re multipliers).

—

Would I recommend this board to others?

If you’re running a 2/3/4-core, and want to use DDR2, then yes.

If you’re running a 6-core and want to use DDR2, then only if you’re willing to: (1) refrain from cranking up the voltage for crazy overclocks; (2) add cooling.

UPDATE: I’ve since put this thing in an Antec 300 case, using a Noctua NH-U12P CPU cooler. There’s an intake fan, side fan, rear exhaust and top exhaust fan. I also JB-Welded a tiny tall heatsink to the southbridge to give it more access to air. Mosfets and Southbridge are both extremely cool now, so I have slightly upped the voltage to get a few more Mhz (still under the 4Ghz mark). Because the components are so cool, I’m not as concerned about the motherboard blowing up when overvolted a little, although I still wouldn’t go crazy. I’ll of course update again if the thing explodes anyway.

If you’re running a 6-core and DONT have extra fans lying around, you’re better off buying an AM3 board and just spending money on the DDR3 RAM – it’ll probably cost just as much as it would for you to buy a bunch of cooling and add it to this board.

If you’re determined to overclock towards the 4Ghz mark, don’t buy this board. Get an AM3 with good heatsinks all around, and a more capable phase power design.

Note that I also tested a more complex render which you can check out in Part 2!

I did a little testing with Maya 2011. If you use Maya and have other machines on your network, you might be tempted to throw satellite on them to speed up renders.

However, you’ll find that this may either help or hinder your overall speed.

I used a test scene in Maya that used mental ray for the rendering. This scene was created by someone a couple years ago in 8.5, and while about 12000 frames in size, most of it wasn’t overly complex. We’re talking images per minute in terms of the output. Obviously, while the render’s being done in 2011, it only uses effects/features that were available in 8.5.

Here are the results. Under the “frames per minute” section, I essentially looked at the file time stamps afterwards and wrote down the number of images for each of the first 5 minutes or so:

—

So what can we take from this?

• The render doesn’t scale 100% with the machines. If the single machine was doing 3-4 frames per minute, you’d think 2 machines would do close to 6-8 frames per minute, but it’s not even close. There’s overhead in distributing the data over the network. You see this evidenced when actually rendering. Using a single machine, the CPU load sits near 100% almost all the time. As soon as you start adding network machines, you find that it idles at least 1/4 of the time (both on the master and the slaves). On simple scenes (talking 10-20 seconds per frame), you’re not likely to see as much benefit as for complex scenes (talking minutes per frame) because so much time is spent at idle just distributing/allocating each frame.
• Never use machines on a wireless network. Note that the peak network data from the master was typically around 5-6MB / second which should have been well within what the N could handle in terms of bandwidth. Wireless seems to bring another issue of some sort that chokes the satellite render.
• Always test the network configuration you’re planning to use against a solo configuration. Remember that my 5-machine test was as slow as the solo configuration.

Why was the 5-machine test so slow?

All I have are theories, but the 2 I’m leaning towards are as follows:

1. Slow machines may bottleneck the render. I don’t know what method Maya uses to determine what chunks to distribute, but if it’s simply a matter of splitting the scene into X equal chunks and sending them off (where X = number of machines or CPU’s), it would stand to reason that a slow enough machine would still be working on it’s chunk while all the others sit waiting. Again, this is just a theory – I have no idea what method Maya uses to determine how much to send to each satellite.
2. There may come a point where the increased overhead per machine added results in diminishing returns.

—

A couple final tidbits:

Note that the CPU load/usage on all the machines jumps from idle to full load during the network render (after it renders it’s piece of the scene/frame, it sits and waits until it gets the next one). This can help keep the heat down (and thus, possibly keep stability up) on all the machines, and keep them usable if they’re being used for other tasks. However, time not spent at 100% load is often time not well spent.

If a machine drops out (goes to sleep, network connection dies, etc), the network batch render just ends. If you have an unstable machine, don’t use it unless you’re able to monitor the render process in case it conks out.

—

In any case, the important thing to take from this is to benchmark before setting off a network render. You don’t have to spend hours doing it (nor should you), but at the very least render a few frames on just the master, and then try your networked configuration. If there’s an improvement, great – go ahead and do the whole thing. If you see a decrease, do the entire thing on the master. Don’t go crazy (like me) and start benchmarking various configurations – you’ll generally spend more time testing than you’ll save during the render process.

Windows cannot be installed to this disk. The selected disk is of the GPT partition style.

The reason is because the Mac OS uses the “GUID Partition Table” partition scheme, whereas Windows/DOS use what’s known as MBR (Master Boot Record). For whatever reason, the Windows installer is incapable of changing this (funny because I used to use Windows installation CD’s on drives I absolutely could not format with other operating systems and it always used to work).

In any case, you’ll need another program to repartion the drive first. I recommend Hiren’s Boot CD (once you go to that site, the link is way at the bottom just below the final ad). You can also use something like SeaTools if you’ve got a Seagate hard drive to wipe the thing clean (or another tool from another manufacturer).

The steps using Hiren’s Boot CD:

• Download the ISO on another computer and burn it the image to a CD or DVD.
• DISCONNECT any other USB drives and other hard drives on the computer you’re installing to so that you don’t accidentally delete stuff on hard drives you don’t want deleted! You should only have the 1 hard drive installed!!!!!
• Boot from the CD.
• Choose “Dos programs”
• Choose “Partition Tools”
• Choose “GParted Partition Editor” (it’s GUI and easy to use – if your mouse isn’t detected for some reason you’ll have to go crazy with the tab button though)
• Select the partitions on the hard drive. You might have 2 showing – a 200MB one (probably shows up as fat32) and the other large one that makes up most of the hard drive’s total size (probably shows up as hfs+).
• –To do this, you click “Partition” – “Delete” – after you’ve done it you need to click the “Apply” button to apply the changes. Remember to do this for all the partitions on the hard drive to wipe it clean.
• Now you should only have 1 item showing in the list – the unpartitioned drive. Click “Device” – “Create Partition Table”
• –The default is named fat or fat32 or ms-dos or something. It’s fine. You may have to “Apply” afterwards.
• Now create a new partition. Choose “Partition” – “New” and select “fat32″ from the side. You COULD choose NTFS, but I prefer to choose FAT32 so that Windows has to delete it and make a new NTFS that’s guaranteed to work. “Apply” again if necessary.
• Now eject the CD, put in your Windows CD, and hit the reset button on your computer.
• When you get to the hard drive screen on the Windows installer, it’ll have an error message at the bottom (where the old GPT message was) about the drive needing to be formatted as NTFS. Select the “Delete” option. Then click next and it’ll automatically format the drive as NTFS and install.
• Once Windows has finished installing, go ahead and reconnect any 2nd/3rd hard drives you had connected before.

I’ll note that ALL the above steps are probably not necessary. However, I prefer to be thorough, and this is the way I’ve done it both times and it worked perfectly.

It’s odd that the Windows installer is smart enough to detect the GPT partition scheme but not smart enough to overwrite it with an MBR scheme. Fortunately, there are other partition managers capable of doing it, and many are included in the Hiren Boot CD. It’s inconvenient, but it works well.

This issue affects the Windows Vista installation disk. It may also affect Windows 7 install DVD and Windows XP installer CD.

Of course I’m unfortunate in that I’m stuck with 2 of these boards. Oh well.

—

The P5KPL-CM is a Socket 775 motherboard, and the biggest problem you’ll come to when overclocking is that the motherboard makes it very easy for the RAM to limit your FSB, particularly with the processors that run at a base of 200Mhz. An example is the E2140 that runs at 1.6Ghz by default (200Mhz x 8).

I’ve got 2 E2140′s, and I tried modding both of them to the 266Mhz boot strap through the pin mod. I used the conductive defroster-repair liquid both times, and tried 3 of the different volt mods to go along with them (in addition to the stock unaltered voltage). At worst, you get a blank screen. At best, you get a screen screaming for a disk with the BIOS and frantically checking your CD-ROM/USB drives for a bios rom.

After the many failed attempts, I did a little searching and found that this is typical of the P5KPL-CM. It just won’t accept a CPU that’s been BSEL modded. My results mirrored what many others have found.

—

Why this is a problem…

The big issue is that at the 200Mhz boot strap, you get 2 options for the RAM aside from AUTO – 800 and 667. This is a lie. The lie is repeated on the boot screen. You’ll need to use Memtest86+ or CPU-Z to see the real values when overclocked. Choosing one of these just picks a ratio for you. 800 is 1:2 (200Mhz –> 400Mhz = DDR2 800). 677 is 3:5 (200Mhz –>  =  333.33Mhz = DDR2 667).

That’s it. Those are your only options. Often, you’ll only have DDR2 800 RAM to go along with your budget board and budget CPU. That means you’ll end up having to choose the 667 setting if overclocking. So how much of an FSB increase can you pull before you get it to 800 speeds?

240Mhz.

240Mhz / 3 x 5 = 400Mhz which is DDR2 800.

Incidently, if you look around, most people hit a wall around 240-250Mhz. It’s the RAM. The motherboard won’t allow a better mem ratio on the 200Mhz bootstrap.

—

Getting the most out of your overclock:

• The ability to change the RAM timings is under the Northbridge Chipset section of the BIOS. 6-6-6-15 is the loosest you can go. Do it.
• Obviously, set the “667″ speed for the RAM to get the most headroom possible.
• Crank up your voltages if need be – WARNING, my settings didn’t fry my parts, but they could fry yours! Be very careful.
  -I cranked up 3 of them to the max setting (there are only 1-2 settings for each). The E2140 I’m using didn’t get very hot (neither did the northbridge with this CPU) so I wasn’t very concerned about maxing these.
  -I turned the RAM setting to 2.0 volts. I did try 2.25V, but that’s dangerously high, and didn’t improve my results (RAM heatsinks got quite hot though – don’t even dare to try 2.25V if you have basic RAM with no heatsink).
• If you happen to have DDR2 1066 RAM around, that’s your best option. The RAM should be able to handle an FSB of 320 (320Mhz / 3 x 5 = 533.33Mhz which is DDR 1066). That said, don’t get your hopes up. Using the 200Mhz boot strap, that’s a fair bit of an overclock on the P5KPL-CM.
• If you have multiple sticks of DDR2 800 RAM (perhaps in another machine…?), grab all your RAM and see which manufacturer clocks the highest. Then mix them in the board as follows…
  My example:
  I had 2x2GB Mushkin DDR800 sticks that worked at 266Mhz FSB (ran at 444Mhz, or DDR2 888 speeds using the 667 BIOS setting)
  I had 2x2GB OCZ Gold DDR800 sticks that failed at 250Mhz FSB.
  What I did was I mixed the RAM. I put the (bad) OCZ Gold in the memory slot *closest* to the CPU, and the (good) Mushkin in the slot farther away.
  I did this on both motherboards, and they both passed Memtest86+ at 266Mhz FSB after this!

• Beg somebody to mod the BIOS. An additional RAM multiplier is available at higher bootstraps. Someone really capable should be able to bring the additional multiplier into the 200 bootstrap (or change what the “667″ ratio actually points to). Alternately, they might be able to fix/disable whatever check is causing the board to crap out when it detects a CPU at a bootstrap that it shouldn’t be at (or they might be able to mod the CPU table so that it thinks all 200-fsb chips are really 266-fsb chips, though this would probably *require* you to BSEL-mod your chip to match.

Those are the only real options. If you’re itching to spend some money, a current CPU will give you the most bang for your buck (you’ll get a higher speed, although you’ll still have limited overclocking). The next best option is a new motherboard that’s more overclocking-friendly (and has more RAM ratio possibilties), particularly if you’ve got a CPU known for really high clocks at stock voltages. The final option is getting faster RAM (DDR2 1066). It’s not great bang for the buck because while the RAM will go higher, the board might choke anyway before you get to the new memory FSB limit.

—

Conclusion

I’m pretty disappointed with ASUS. This motherboard really should been labelled under their ASRock line. Even ECS boards (which I’ve had many of in the past and which all sucked) can apparently handle the BSEL mods. I stopped buying ECS a long time ago, and based on the last few ASUS boards, I’ve been buying more MSI and Gigabyte lately. ASUS just has too many boards with poor overclocking options, and they’ve gone from making a lot of great enthusiast motherboards to a lot of mediocre motherboards. It’s a shame when a processor known for insanely easy overclocking is barely able to get a bump on an ASUS board.

In any case, BSEL won’t work, and while you can toy with the few settings you’re given, if you want to hit the high Gigahertz, you’ll either need great RAM or an already-fast CPU. Then again, you might just want to buy a quality motherboard from another manufacturer.

—

UPDATE: I was able to get to get a little farther – 285 Mhz FSB on both motherboards by disabling the onboard video and using dedicated video cards. Both were stable although I blew a power supply. 290 Mhz failed Memtest86+ although I’ve got both Mushkin sticks in the 1 machine now and am looking to see if the board will go higher.

Note that the RAM gets extremely hot at 285Mhz (DDR2 950 speeds). I have a dedicated dual-fan RAM cooler on it (the built in heatsinks aren’t enough), as I wouldn’t want to keep pushing the clocks at the temps they’ve been sitting at.

Incidentally, the OCZ sticks got the hottest by a long shot (even though they were closest to the CPU fan and would have received residual airflow). I’m surprised they didn’t melt. The power supply that fried was an OCZ also. I’m quickly losing faith in OCZ.