MSI 850 Pro2

One of our guests...

Whoever thinks there isn't a difference between a Pentium 4 and a Pentium 4 is wrong. Whoever may think that all Pentium 4 mainboards are created equal is also wrong. We are in the lucky position of being able to prove both of the statements above, as we have spent a few days together with a couple of P4 processors and motherboards. Out of the four, we will be mainly focusing on one of the motherboards - MSI's 850 Pro2 mainboard, built on a 4-layer design for Socket-423 Pentium 4 processors (aka Willamette), with the i850 chipset. Without giving away too much of our article, we can point a few things out, namely, that MSI is the first with a 4-layer P4 board and that is supposed to drive costs down. The 850 Pro2 also has a spectrum of features that the original 850 Pro didn't, but the ASUS P4T did... and since one of our P4 CPUs was from the "appropriate" batch, we were actually able to put these features to use.

What makes a Pentium 4 processor so good then? There are some general characteristics of P4 systems that make them "good", while there are several unique tricks that are not offered by everyone. Let's discuss the first type first, only very briefly. The most obvious thing here is FSB and bandwidth, which we have thoroughly discussed in our last Azza 633X-AD review. To put it simple, almost all recent platforms are limited by their FSB and/or memory interface. VIA's Pro266 chipset is trying to match DDR memory with an SDR FSB - the worst choice. VIA's KT133A has an SDR memory interface, but a DDR system bus (EV6). The new DDR chipsets for AMD processors seem to be much more balanced as they are offering a completely "Double Pumped" platform, as both the memory and the FSB operate at a double data rate (DDR). Intel tackled the problem in a major way with its i850 chipset and RDRAM support, while they did introduce a few "quirks" with their new CPU...

Pentium 4 basics

You will have to get used to this, but once something is invented in the computer business, it will later start to scale exponentially. You will always have double the same thing, then four times as much and so on. You can think CPU, AGP, memory speed or almost anything here. Remember the original 486DX CPU at 33MHz? Then we had 486DX2 at 66MHz and we didn't need much time to meet the - quite illogically named - 486DX4 CPU at 100MHz. The same happened with AGP 1X, 2X, 4X and now 8X, and the same is happening with the system bus, or FSB, for short (Actually, the 486DX2 is a perfect example here as it was the first CPU that had to separate its core frequency from the external bus. Before, the CPU and the bus were operating at the same speed, but with the introduction of a 66MHz, blazing fast processor :), that was no longer possible. So, the bus remained at 33MHz, while the CPU was using a double clock, internally. Nowadays, this is normal, as all CPUs come with a multiplier, which multiplies the external bus (FSB) to get the internal processor speed.). Only the scaling of the FSB is different, as clock rates are maxed out at 133MHz (as of yet), but manufacturers are pumping more and more data using that clock. More precisely:

• Current AMD platforms are all using the EV6 bus protocol licensed from the Alpha guys and operate at 100 or 133MHz, but use a DDR technique. So effective throughput is 200/266MHz if we translate that to a conventional SDR bus, but it is important to understand, that Athlons/Durons still operate their FSB at a 100 or 133MHz frequency!
• Intel is now doing the same for the first time, with Pentium 4. The FSB is operating at 100MHz, but data is sent a QDR (Quad-Data Rate) fashion, which means that four units of data are transferred with each clock. This is the same as if we used a 400MHz SDR bus, but the P4s still operate their FSB at a 100MHz frequency!

Most of you have known this for quite a while, but for the sake of completeness and a better understanding of what's coming, we wanted to make sure ;-). Marketing terms such as "unparallelled 400MHz FSB" can also be put to rest this way. Now for the Pentium 4 CPU and RDRAM: as with every CPU, the Pentium 4 also uses an internal multiplier to derive its final speed. Both of our 1.4GHz P4s use a 14X multiplier, which makes sense knowing that the FSB is running at 100MHz (if it did run at 400MHz, as you see on most mainboard boxes, 14 x 400MHz would give us a 5.6GHz CPU... not bad :-)). RDRAM is doing just the same: PC800 RDRAM is using a 4X multiplier AND is double clocked (so 4 x 2 x 100MHz = 800MHz), with PC700 and PC600 using 3.5X and 3X, respectively. As the external bus is 64 bits wide, the maximum usable bandwidth in the system is 8 bytes x 4 x 100MHz = 3200MB/sec. RDRAM has a 16 bits wide interface, but the i850 chipset is a dual-channel RDRAM implementation, so bandwidth with PC800 RDRAM is 4 x 2 x 100MHz x 2 bytes x 2 = 3200MB/sec, which just matches the FSB bandwidth. The upcoming nForce 420 (aka Crush 12) chipset from nVidia is using dual-channel DDR SDRAM and has a memory bandwidth of 4200MB/sec with PC2100 modules, but the system bus remains at 2.1GB/sec making for an unbalanced system (in reality, the most of that 4200MB/sec is needed by the integrated graphics). The balanced and very high maximum bandwidth is probably the strongest point of the whole Pentium 4 story.

We could take the route of starting to compare the P4 with the Athlon now, but we won't. To put it simple, there are some serious downsides of the P4 that most believers don't like to talk about. The die size of the current Pentium 4 CPUs is huge, 217 square millimeters, to be exact. That means P4 is running hot, it is a rather costy thing to produce and also implicitly means that there are some features of P4 that could have been implemented, but haven't. The FPU part simply sucks, with only one castrated pipeline feeding data - sure, there is SSE2 to compensate all that, but once it comes to pure FPU tasks, P4's almighty bandwidth isn't going to help much. L1 cache sizes are also a bit small (although they are FAST), with data cache being only 8KB and no code cache. The double pumped nature of the ALU (integer part) helps to hide the incredible penalties certain instructions generate, but surely doesn't take performance to new dimensions, like Intel claims. We sincerely hope that Intel will manage to repack all the stripped features once Northwood is released (and hope that the talks about 512KB L2 cache will prove false, as more L2 cache is NOT the best way to raise P4 performance now... integrating 512KB L2 cache eats up all available space for additional pipelines, etc.). If they do, AMD followers (and ASM coders :-)) will have a harder time trashing the P4. This is about as much as we wanted to say about performance compared to other solutions (read: AMD) and from now on, we will be focusing on the P4 hardware we had.

Extreme overclocking / I

Motherboard requirements

Overclocking Intel CPUs became somewhat difficult when Intel introduced the locked multiplier scheme for its processors. Difficult, but not impossible, as the final CPU speed is always multiplier x FSB and FSB is set by the mainboard and not the CPU. The CPU does tell the mainboard what FSB it is supposed to run on, but motherboard manufacturers have the choice of adding additional adjustable FSB speeds and usually they do. The story is exactly the same with the Pentium 4 - the multiplier is fixed, but mainboards have the chance of offering higher FSB speeds than 100MHz, which allows us to overclock our shiny new P4 to crazy limits... does it? Not exactly, as the P4 systems introduced a few new overclocking "factors". The system bus is quite happy at even 133MHz, that's no problem. RDRAM, however, feels quite ill at anything beyond a 10-15% increase. So, at a 120MHz FSB, you will surely get lockups not because of the CPU, not because of the GMCH failing, but because of RDRAM not being able to tolerate the 8 x 120MHz = 960MHz frequency. The CPU could also be a limiting factor, the point here is that RDRAM will always max out at about 110-115MHz, regardless of the CPU.

Luckily, there is a way to get around this, as PC800 RDRAM is not the only choice for P4 users. PC600 RDRAM is also available and all mainboards should be able to cope with them. All they do is that they adjust the multiplier of the RDRAM according to the information given to the board by the memory module. PC800 RDRAM uses a 4X multiplier, whereas PC600 needs 3X and since RDRAM multiplier is not locked, any mainboard capable of overriding the specified value and allowing the user to select between 4X and 3X has a huge "+" on the overclocking features list. Let's suppose we select a 133MHz FSB. This is something PC800 RDRAM definitely cannot handle, as 4 x 2 x 133MHz = 1066MHz is way out of range. On the other hand, if we set the RDRAM multiplier to 3X as if we had PC600 RDRAM, we would get 3 x 2 x 133MHz = 800MHz, which is exactly the spec for PC800 RDRAM. Adding the additional 10% tolerance most RDRAM modules have, FSBs of 145-150MHz are achievable this way! Let's sum up our findings. Every P4 mainboard expected to overclock well should have the general overclocking features:

• Wide range of FSB speeds, possibly with 1MHz increments,
• CPU Core Voltage (VCore) adjustments, at least up to +0.1V,
• Possibly RDRAM voltage adjustments (similar to I/O Voltage, or VIO, for short).

This would allow us to reach 110-115MHz FSBs if the CPU and RDRAM is right, for sure. If we wanted a true overclockers board, there is one additional feature that HAS to be there to unleash FSBs between 115MHz and 150MHz:

• Manual selection of RDRAM multiplier between 3X and 4X!

Currently, there are two mainboards that support all of these features: the ASUS P4T and the newly released MSI 850 Pro2. The first is the one and only ATX P4 solution from ASUS, it is based on a 6 layer design, doesn't come cheap, but has been the all-time favourite for P4 users due to its capabilities. The latter, MSI's newcomer is supposed to change all this, as it bears every feature of the P4T, but comes with a reworked 4 layer design and noticably cheaper. Both boards support Socket-423 (Willamette) P4 CPUs, up to 2.0GHz and if you seek to seriously overclock your P4 processor, you currently have no other choice than these two offerings. Not all Pentium 4 mainboards are created equal, remember?

Extreme overclocking / II

Processor requirements

Okay, we have one of the boards above (we actually did, out of the four hardware we mentioned, two were these two mainboards) and are ready for some serious action. We plug in our fresh and new 1.7GHz P4, set the FSB to 133MHz, scale back the RDRAM, save the settings in the BIOS and reset... and... and... black screen. Okay, we reset the CMOS, dive back to the BIOS settings and start scratching our head: what happened? The solution is simple. 17 x 133MHz = 2260MHz, which is simply too much for a 1.7GHz part. This means we have to experiment with a lower clocked P4 to achieve better results. Why not try with a 1.4GHz part? Sure. 14 x 133MHz = 1866MHz, which is also way too much for most 1.4GHz P4s. Most, but not all! What is the differentiator then? To put it simple, if we somehow received a 1.4GHz P4 that in reality had a 1.7GHz core, but was labeled 1.4GHz, for any reason, we would be almost sure to reach 1866MHz. Sounds great, but why would anybody, especially Intel, label a 1.7GHz part for only 1.4GHz, when 1.7GHz makes them more money? The answer is simple and Intel has been doing this ever since they existed (remember those SL2W8 P2-300MHz CPUs, that could do 450MHz with default voltage?).

As time passes, the production of a certain CPU surely yields higher output. Production lines get better, new steppings are introduced, etc. Stepping is the key word here. Stepping stands for the revision of the silicone used in the chip. Higher stepping number/code usually means that a few things have been patched, produced heat has been lowered and yields/frequencies are better. Think of it as a new BIOS revision (ie: for a mainboard). But what happens when a newer stepping allows Intel to produce all P4 CPUs at or above 1.7GHz, but there is still serious demand for lower speed parts? Should Intel get back to older production lines? No way, costs would raise without any benefits! Should Intel lower the price of the 1.7GHz part all the way down to 1.4GHz, for example? For us, that's a yes, but for Intel, certainly no, as they would lose a lot of money from people who would have bought the 1.7GHz part for the original (high) price. The only - profitable - solution is to satisfy demand with 1.4GHz chips, even if all chips are 1.7GHz by nature :). So, the multiplier lock kicks in and a CPU with a newer stepping and a "could-have-been" 1.7GHz core is released, at 1.4GHz, with a 14X multiplier. These are the CPUs we are looking for! They cost you "only" 1.4GHz, but are 1.7GHz inside and will probably run with our new boards at 133MHz FSB at 1866MHz. Take a look at the table below, taken from Intel's site:

One immediately notices, that there are only two steppings for all CPUs, one being B2 and the other (newer) C1. If you recall, P4 was originally released as 1.4 and 1.5GHz, with 1.3GHz following LATER! The B2 stepping is only available up to 1.5GHz and as it is the older one, it is safe to guess that original P4 CPUs were based on this core. Thus, buying a 1.3GHz part (in case it's cheap) makes perfect sense as it is from the same stepping as the original 1.5GHz CPU. It was even introduced later, probably for demand reasons, so being able to run it at 1.5GHz is almost certain. Please understand, that overclocking isn't independent of luck, as chips from the very same batch with the very same stepping all overclock differently. But not very differently. If someone has a P4 1.4GHz running at 1.9GHz, it's not because he was so incredibly lucky, but because he has a CPU with the newer stepping that was downgraded by Intel AND he was lucky to get a good part from those. Crazy high overclocks (let's forget extreme cooling methods now, with a better stepping, they also do better) always attribute to "tricks" like this.

Which one is the right P4 to get then? Well, just take a peek at the C1 stepping! ALL current P4 processors are available in the C1 flavour and man, what a flavour that is! Even the 1.8GHz part is using C1. Our theory is partially proved here, as Intel was fine with producing 1.3-1.5GHz CPUs with B2, but for some reason they later introduced these CPUs with C1 as well (B2 is probably no longer being produced, at all). Does this mean that a 1.3GHz P4 labelled SL5FW has a 1.8GHz core, for sure? Probably yes, although production does work by sorting the CPUs in categories after they have been pre-tested by Intel. As always, there are some faulty ones and others just don't make it to 1.8GHz and have to be downgraded. But in the long run, C1 stepping CPUs are all based on the highest core produced with that stepping. To make sure you are in the "long run", try to select a CPU with the correct sSpec AND the highest production week possible (more one this later).

Our 1.4GHz CPUs

If our theory is correct, we have to be able to run a C1 stepping 1.4GHz part at 1.7-1.8GHz without raising the core voltage AND we shouldn't be able to run a B2 stepping 1.4GHz P4 at 1.7-1.8GHz, at all! Whoever thinks there isn't a difference between a Pentium 4 and a Pentium 4 is wrong, remember? We had a SL4SC (B2 stepping) and a SL4WS (C1 stepping) 1.4GHz Pentium 4 to experiment with. Below is the highest speed we could achieve with both, along with the stepping info:

B2 stepping and a modest overclock here...

...whoa, C1 stepping does look better!

Our C1 based P4 ran at 1.75GHz at default core voltage, something a truly overclocked CPU definitely cannot do. Our guess, that such 1.4GHz part bears a hidden 1.7-1.8GHz core seems to be proven, although 1866MHz was as high as we could go, with any voltage (eighter we had bad luck within the right batch, or the board isn't stable at above 133MHz FSB). Is there any point in buying a B2 stepping CPU when you have a choice? No. They (should) cost the same as resellers (should) only care about the stock speed.

"Close encounters of the first kind"

In the third row on the picture above (below the sSpec SL4SC), you will see the code L110... The first digit/letter stands for the production place, L=Malaysia in this case. The second digit is the production year, surprisingly 2001 here. The next two digits, in the 3rd-4th place are the production week, ranging from 01 to 52. Remember, out of the same stepping CPUs, higher week number is usually better as production is always ramping up with time.

Rumour has been on the 'net saying that the P4 has a thermal protection circuit (it does) that kicks in from time to time and throttles the clock of the CPU down halfway. In other words, it can happen while playing Quake that the 1.4GHz CPU overheats and the integrated circuit lowers the clock speed down to 700MHz! Intel engineers denied these rumours and claimed them to be completely false (although if you run your P4 without a fan this feature could kick in). We haven't seen this feature turn on once during our testing period and SiSoft's Sandra says this feature isn't even enabled:

No throttling here...

Therefore, we (also) consider this rumour to be completely false.

The first 4 layer board

Being first is a cool thing and recently MSI is quite used to the situation. From an engineer's perspective, designing a 4 layer P4 board in not an easy task and thus MSI has every right to be proud. From the customers perspective, however, layers don't matter at all. Price does and features, as well. "Luckily", the two things are directly connected, as 4 layer boards should be cheaper than their 6 layer cousins as they are easier to produce (not to design). We asked MSI about this and they did in fact reply that the 850 Pro2 board should be 5 USD cheaper than the 850 Pro, even at introduction. While 5 bucks don't sound much, keep in mind, that the 850 Pro2 is a freshly released board and is surely carrying a price premium because of this. The 850 Pro was one of the cheapest P4 boards anyway, so a full-featured and new mainboard coming at the lowest price is nothing to be ashamed of. In addition, the ASUS P4T, the only other board with the same excellent features is on the most expensive side. If stability and speed is right, MSI for now has an absolute winner, no question about that.

We have pretty much discussed all that's in the table above, but it's a nice thing to have things summed up. We will not be touching MSI's proprieraty features, like PC2PC USB pseudo-networking and D-Led technology, as they have thoroughly been discussed everywhere, even in our own review of the MSI K7T Turbo-R. The 850 Pro2 comes in a (nice) standard MSI box, with all required cables, heatsink retension clips, PC2PC USB cables, drivers and manuals inside. The manual is pretty decent, although under the BIOS Settings part we still see options like "this sets the RDRAM Turbo option. Options are: enabled, disabled". Probably anyone can figure that out by himself :), a brief and real description would have been nice... (you can find similar "descriptions" in any manual, not only MSI's... in fact, this option isn't even a real one in this case, we just wanted to display an example that there is still place for manual improvements for any mainboard, any company).

Click the box!

The board is equipped with a few 1500uF and 2200uF capacitors. At first, they didn't seem "enough", but the whole mystical belief around capacitor number and size directly correlating with stability and overclocking performance is a bit overexaggarated. Ask any engineer, it is a bit more complicated than that :). Nevertheless, MSI's offering turned out to be an extremely stable one, matching the ASUS P4T exactly with the highest achievable FSB speed.

Stability factors

Overclocking fans will be happy to see that the North Bridge (oops, better said GMCH) is equipped with active cooling. The i850 GMCH does tend to run hot, but we were more statisfied with the huge passive heatsink we saw on almost all other P4 boards, as it does its job perfectly, but is much more quiet. For a short testing period though, the active solution is better, as the GMCH isn't to overheat, even with "insane" FSBs, for sure. The South Bridge (ICH) is the standard UDMA/100 capable ICH2, with 2 IDE channels supporting up to 4 IDE devices.

Click the fan!

The 850 Pro2 stops the tradition of equipping high-end mainboards with AGP Pro slots, or at least, does not bear one. On the other hand, the AGP 4X slot is enforced by the AGP retension mechanism we first saw on Gigabyte boards. It is a nice thing as it makes sure the AGP card is firmly seated and doesn't move once it is installed.

Rev:1, still works fine

Tests and results

Our first Pentium 4, with the mean B2 stepping was a boxed one, therefore, we used Intel's own heatsink+fan and the 2 pieces of Samsung 64MB PC800 non-ECC RDRAM for our tests.

Click the RDRAM chips!

To tell the truth, after seeing Kyle's review at HardOCP about the P4 and TBird shootout, we were almost sure that our magical 1.4GHz C1 stepping P4 will take us to above 2GHz. Why? Kyle had a 1.7GHz CPU which he overclocked to 2.1GHz and our 1.4GHz one was even younger (produced later) than his, but with a 1.7-1.8GHz core. Our hopes quickly vanished when the MSI board refused to do anything at a 135MHz FSB, while we would have needed 150MHz to reach 2.1GHz. At first we were left with a bitter taste in our mouth (I guess HardOCP isn't Hard OC(P) for nothin' :-)), but later came up with the idea of getting an ASUS P4T as well, just to make sure the MSI board isn't holding us back. Fortunately or unfortunately, we replicated our results with the P4T and 133MHz was the absolute top there too. In fact, having used the P4 system for almost two weeks now, we have to lower our results by another 3-4MHz. 130MHz is where both boards end (with the CPU being truly stable, not just for benchmarks), but since Anand of AnandTech was able to run his engineering sample P4 in an ASUS P4T at over 140MHz FSB, it is a safe thing to conclude that the CPU is holding us back in this case.

The P4T did have a few additional features that we found quite amazing, one is the metal backboard that you can attach to the mainboard, fix the heatsink to the backboard (and not the ATX 2.03 ready PC case) and later screw the whole thing in any ATX case, like with any older mainboard. The other was a BIOS feature - if you overclock your CPU to an extent where it fails to boot, the P4T recognizes this and without you touching any INSERT keys, resetting the CMOS or any other hooplah, it just resets the computer and the overclocked settings, enters the Setup for you and warns you that your CPU has failed to initialize! Quite elegant, we must add. Sure, even the P3B-F had this feature, but we are still amazed when we see it kicking in. Apart from these extras, however, the P4T had an interesting role in our tests: it didn't prove to be more stable than MSI's offering, while being significantly more expensive. All in all, this is an example of mainboards maturing with time... The P4T is simply obsolete (for the price) now and the 850 Pro2 is clearly the P4 board to have. It has all the features and is cheaper. In the meanwhile, ASUS is preparing its P4T-F with a 4 layer design, so the current situation might change (and to be fair: the P4T simply blows away the original 850 Pro.).

However, until ASUS, Gigabyte or anyone challenges MSI, the 850 Pro2 is all safe from any competition and MSI can sit back and relax knowing that they are on rampage :-). Being first does matter, I guess, just ask their Sales Department. Back to the tests (we are almost done, however is still with me: thanks, I know it's a bit "too long"): one measurement we were determined to make was the maximum tolerance of the RDRAM chips, in other words, the highest achievable FSB speed without scaling the RDRAM multiplier back to 3X. This turned out to be 115MHz with the 850 Pro2 and our B2 processor and we got some INCREDIBLE results. The ASUS P4T actually fell behind MSI - even with RDRAM set to Turbo - by quite a margin, something we were not used to in our tests before.

Now THAT should do it...

If we set the RDRAM multiplier to 3X, we could overclock the FSB from 115MHz to 118MHz, where our B2 stepping P4 simply maxed out. It is not hard to guess that the additional 3MHz in FSB is simply not worth the loss in RDRAM speed, but just to make sure noone argues, here are the results:

Not enough

The true ride however is with our C1 stepping P4, for sure. Although we weren't able to match RDRAM scores at 115MHz FSB with the default 4X multiplier (133MHz x 3 x 2 is the default 800MHz, whereas 115 x 4 x 2 = 920MHz... the 133MHz x 4 x 2 = 1066MHz is a no-go, of course), the increase in FSB and CPU speed greatly compensated for all this. We won't be giving you any Quake, 3DMark and Unreal scores as we have reviewed at least 3 different P4 combos with GeForce3 video cards in the last month and the goal of this review was to find the place the 850 Pro2 amongst other P4 solutions. The Sandra scores are here, however :)

This, we accept :)

"Whoever thinks there isn't a difference between a Pentium 4 and a Pentium 4 is wrong. Whoever may think that all Pentium 4 mainboards are created equal is also wrong." - You have seen appropriate examples proving both statements above. The MSI 850 Pro2 board is one of the participants that are the cause these inequalities, as it simply trashes the competition in the price/performance battle. It has EVERY SINGLE feature that is needed for a top quality Pentium 4 mainboard and is cheaper than anything else currently available for the P4, not only its main competitor, the P4T. The other thing we learned to look out for is Pentium 4 sSpec information, as a CPU from the right batch can take you to 200-300MHz higher, with ease. To put it simple: if you are planning on investing in a P4 setup now, the 850 Pro2 is the board to get and a C1 stepping (lower speed) Pentium 4 is the CPU to plug in that board.

(Marton Balog, alias Parci)

This is the first time we see MSI truly standing out of the competition, no matter where we look at it. The 850 Pro2 is the first 4 layer board on the market. The 850 Pro2 is the cheapest available P4 board. The 850 Pro2 is right there with the ASUS P4T regarding features and both boards are miles away from competitors. It is without a question that until someone successfully challenges MSI, the 850 Pro2 is the worthy one for our PLUS A LOT Golden Award. Congratulations!