Coolers for Pentium 4 `2002 Q3

Introduction

It is accepted that Intel Pentium4 processors do not pump out heat that much and do not require special cooling devices - the cooler that comes with the company's BOX option suffices. Indeed, with the vast variety of coolers for AMD processors the list of coolers for Intel counterparts looks scarce.Therefore it doesn't take much to review practically all the popular makes.

To pick up the coolers, we popped about all the near computer shops and lucked out to get 4 aluminum coolers and 2 copper ones. A bit short of this, moreover, the copper ones, as the shop-assistants said, were completely out of demand. Any other coolers but for those in the BOX make neither can boast consumers' popularity. What's up? Pentium 4 is about the same in theoretical heat emission as Athlon XP. Let's sort this out in detail.

First off, there are Pentium4's based on the old Willamette core and on the 0.18 mk process technology as well as new Pentium4's based on the Northwood core (0.13 mk). The difference in heat emission between the two cores is rather big: the Willamette running at 2 GHz pumps out over 70 watts of heat, while theNorthwood-basedprocessor emits merely ~50 watts. The latter has much more transistors (512k cache size vs. 256k for the Willamette) and runs at much higher clock speeds.

Even the 50 watt is too much. Meanwhile, you can come across some messages online saying that Pentium4 stays operable even with no cooler at all.As a proof, they brought in that famous video clip showing a system running Quake 3 with the cooler removed. Pentium4's, Pentium3's and a couple of Athlon's were run through such tortures successively. The Athlons burnt down completely, the Pentium3 hanged, but the Pentium4 survived and was still running.

The secret of the Pentium4 is very easy to reveal - a quick look at the processor is enough.

The processor core is shielded with the so-called IHS (stands for 'Integrated Heat Spreader'), a copper plate several millimeters in thickness. The copper plate is nickel-coated. Such design has several advantages. First, local overheats are prevented, since some processor blocks are heavily loaded (e.g. some blocks run at a doubled clock speed). Heat emission of such block is rather high and could have resulted in a failure of the whole processor. The copper plate prevents this through distributing heat all over the surface.

The second advantage is that the area of the copper plate is twice as great of the core and thus does the same as the copper base on coolers in AMD processors.

Finally, the copper plate does the protection job - it preserves the delicate processor core from various damages (e.g. core chipping while installing the cooler).

However, a question is still there:"if we remove the radiator completely, would the copper plate of only 2 mm in thickness be able to pump out heat from the running processor?". Definitely not. The thing is that in such a critical condition the protective mechanism of Pentium4 comes into play. The outcome is evident - the processor performance drops sharply. How exactly does that occur? There are various opinions on that. Some say that the processor clock speed decreases as a result of the multiplier inner changes. Others assert some cycles are missing. In any case that doesn't matter much (for more details, browse the Internet if you are curious about that) - most importantly, under any cooling the processor stays operative.

By the way, you can see in that movie how the protective mechanism works. Note that in the frames per second reading is displayed in the corner of the screen. After the cooler was removed, the figure dropped from 200 to 25 frames.

On completing the tests I decided to do my own research of that process. I didn't go too far - I simply switched off the cooler fan. Then I observed the temperature rise and from time to time took the Sandra scores. As soon as the temperature reached 80 C, the following results were produced:

so I thought it would be reasonable to stop the experiments.

That mechanism of thermal protection has its own drawbacks. Suppose you assembled a Pentium4 1.6A system and set the bus speed to 133 MHz. In most cases the system will run absolutely trouble-free. But interesting begins afterwards!As we set the speed to 150 MHz, the system failed; we pushed the voltage from standard 1.5V (supposedly) up to 1.65V and the system is back on track, running smoothly. Why not to overclock it more? And we install a bus whose speed is over 150 MHz, raise the voltage to 1.75V and higher - all is running OK. We seemed to have finally arrived at the overclocker's paradise - the processor temperature in the Windows OS did not exceed 50-55 C, the clock speed boost was as high as 1000 MHz, and the performance boost in the tests was adequate to the clock speed boost (~20%). But interesting occurs afterwards - you start a 3D game and in about couple of hours fragging monsters you end up noticing (if lucky) that the game speed is going down.

In fact, I brought in the ideal example - reality is rich in more varied options. In every specific case numerous factors matter, starting with the choice of thermopastes (by the way, it is one more review of ours ), current voltage Vcore up to the processor stepping. The main conclusion can be formulated like this - a high-powered cooler is needed for substantial overclocking. Therefore, we arrived at the thought that copper coolers are anyway necessary for cooling Pentium4 processors.

The last what I wanted to say in the theory section is this: you'll never come across processor overheat in the Windows OS while running software which is not hungry for resources (office programs, Internet applications). It means - when not loaded, a Pentium4 system switches the processor to the low power consumption mode, which results in the temperature drop (while writing this text on my PC, the temperature of the Pentium4 1.8A-> 2.4B x 1.65V is not higher than 42 C).

Most interesting is that Athlon XP practically never switches to the standby mode. Even if there are no active tasks running, it is working at its full power emitting lots of heat. To activate the standby mode, some software like VCool or a combination of Wpcredit + Wpcrset is needed. They enable the chipset's function which disconnects the processor from the system bus.

So that all the above did not look like criticism towards AMD, let me demonstrate that at the full load Intel processors heat up no less than their competitors. For that, I compared the maximum operating temperatures of the following processors:Athlon XP 1600+ and Pentium4 1.8A

coupled with a TT Volcano 7+ cooler

Therefore, we found out that cooling Pentium4 processors is as important a task as for AMD processors. But with one reservation in mind - I mean the processor will be overclocked, since the standard cooler with the standard thermal interface is more than enough for the processor if it is running at its rated clock speed. On the other hand, Pentium4 clock speeds keep rising (by the end of the year 2.8 GHz or even 3 GHz processors will be launched). Our research is going to be quite topical, since use of such powerful processors implies immense heat emission. Mind you, nobody is willing to pay a lot for the fastest (and thus most expensive) processor which ends up losing its performance because of the inadequate cooling or a weak cooler (costing no more than $10) :)

A couple of words on the measurement procedure. As is the case with the recent review ("A Roundup of Noiseless Coolers for the SocketA"), I will rank the most subjective parameters with scores, i.e. the higher the score, the better that parameter is. The subjective parameters are first the cooler's noise level, ease of installation and further removal.

But the prices and temperatures of the processor are precise enough and will be given in figures. The processor temperatures and fans' rotational speeds were measured with Motherboard Monitor v5.1.9.1. The test setupwas assembled in an Inwin S506 case with the cover removed (for a better perception of the noise :). If the cooler had related thermal interface, then the latter was removed and further measurements were carried out with the silicon-organic paste KPT-8 (the cheapest of all the tested). The processor was heated up using CPUBurn, with Pentium4 1.8A used as the tested processor (all Pentium4 processors have thermal sensors integrated in the core). For increased heat emission we set its bus speed to 133 MHz (which totally gave 2.4 GHz) and raised the voltage Vcore up to 1.65V.

While describing the cooler I'll draw your attention to the fastening convenience. Also don't forget about such factors as the fastening reliability, radiator's dimensions, material, appearance, processing quality etc. Most of these are fairly subjective (especially, the appearance :), so I'll try to be as impartial as possible through presenting more illustrations.

Thermaltake Volcano P4

Let's take a look at the cheapest Thermaltake model (by the way, it's the most common - you can find it at practically every computer shop).

The design is classical - it is made up of an aluminum radiator with a fan on top). The cooler is easily fastened to the socket with clamps (rather, to the plastic frame)but is more difficult to remove - every clamp has to be turned up separately and levered up slightly with a screwdriver.

The cooler sells in an ordinary cardboard box with nothing else inside but the cooler itself.

The cooler is heavy enough. Most of the mass is taken by the aluminum radiator made up of a couple of dozen fins of equal height.

Attached to the radiator using two screws is an iron frame with a 70 mm fan (the dimensions 70x70x15 mm) on top. The quality of the fastener is very high, so no vibration is produced.The fan's rotational speed is about 4800 rpm, which produces a 30 cfm air flow. In so doing, the cooler is rather noisy - the noise level is 37 dB as per the manufacturer's specifications.

the mean time between failures is 50000 hours

Look at the cooler base - it is processed fairly crudely. In the center there is a rectangular pad of the Bergquist 225U thermal interface. Its heat conducting material has made a very good showing.

Summing it up, the price for the cooler is not high at all - around $8 a piece.

Elan Vital FSNW01-FC

Well, let's examine the cooler more closely.

The appearance is not impressive at all - the cooler is light (weighs only 243 g), small (the dimensions 69.1x83x55.8 mm) and the fins are very thin. A smaller fan of 60x60x11 mm dimensions adds to the mean appearance.

the maximum rotational speed is 4800 rpm

Have a look at the cooler from the side.

The fins are very thin and under any temperature they stay cold. The cooling system is absolutely inefficient.

The cooler base is also not impressive - the processing quality is higher than in Termaltake Volcano, but grinding traces are anyway visible. Originally, there was a thermal interface of the base and then it melted down early after the first test and firmly glued to the processor. All in all, I did not find any pluses in the cooler, although the Elan Vital has a trump card in stock - the rotational speed varies with the processor temperature. This function was expected to reduce the noise level. As per the manufacturer (by the way, the design of their web-site has been renewed :), the noise level is not greater that 38 dB (depending on the rotational speed). The factual noise level is lower than in Termaltake Volcano,but all the same is rather high.

In the end, despite the much-vaunted noise-reduction technique the cooler did not appeal to me.

It has only one merit - it is very easy to fasten. For that, there are two metal clamps that press the cooler firmly. It is also easy to remove. around $8 a piece.

Standard Intel cooler (Intel BOX)

The user doesn't have to care about that, since it is just coolers supplied with the Intel processor in the same boxhave always been of superb quality.

No exception was this time - the cooler manufactured by Sanyo made a very good showing.

The cooler is very easy and fast to install. It suffices to snap shut the fastening frame at the limiting frame of the motherboard and turn the levers smoothly (shown with white color on the photo).

The cooler is not that easy to remove, though - you've got to use a wide-bladed screwdriver.

If you look at it from the side, you can see that the cooler base is thicker in the center. On the base there are over twenty fins of varying height.

A plastic frame with the fan is installed on the base. Unlike other coolers, it is fastened to the radiator without any bolts or screws (like it used to be in the ancient times when they built without nails - here no screws are used :). This fastener turned out to be very strong and reliable - no slightest signs of vibration. The fan was really attractive - first, it changes the rotational speed with temperature. You can't see the thermal sensor - it is well hidden inside. All this has resulted in very low noise levels - in my opinion, it is the quietest cooler among the reviewed.

The contact of radiator with the processor is designed in a tricky way. First, there is a thin foil layer on the radiator and its area is slightly greater than the processor area (the foil is easily seen on the photo). Secondly, a layer of thermal interface is applied on the foil (shown in black - not visible in the photo). After second or third removal of the cooler the thermal interface layer destroyed and I cleaned it off. This thermal conducting material is not bad at all in quality, with the temperature difference compared to the KPT8 paste was 2-3 C in favor of the latter.

Finishing the description I say that I have seen boxed coolers of different radiator shape. They were also made by Sanyo and I think the test results would be the same.

Thermaltake Dragon

The cooler sells in a typical of Thermaltake transparent plastic package. The cooler comes dismantled:the radiator, the fan and the plastic pressure plate are all separated. Besides, the package has an adapter to plug the cooler directly to the power supply. It's a pity no other device for manual regulation of the fan's rotational speed was there in the package.

The cooler is very easy and fast to install: the radiator, the fan all are fastened by the pressure plate with two levers. In the installation the levers are resting on the small pads and anyway they need to be turned uniformly and precisely.

The cooler design is a clone of the Intel reference cooling system. But the clone is not absolutely identical - Thermaltake engineers modified the bending direction of all the fins into the reverse. The fin bends are impressive - each fin is smoothly bent in all three dimensions.

Why is that needed? No doubt the developers pursued a good goal in mind - to make fins fully matching the airflow direction. This could have resulted in the radiator maximum efficiency with a very quiet operation of the entire cooler. But in practice both goals have failed to be met. The cooler efficiency does not surpass essentially that for aluminum coolers, and noise is hard to stand. The maximum noise is produced just as air passes through the radiator fin and is not caused by the high-speed fan.

Additional noise may arise because of the loose connection between the fan and the radiator. Therefore, when inadequately fastened, the cooler may produce excessive vibrations.

the maximum rotational speed is 6000 rpm

The fan is placed into a steel housing badged with the Thermaltake logo. I plugged it separately - of course, the noise was there, but easier to put up with.

If you look at the radiator more closely, you will see that the base is of copper with the fins made of aluminum and simply painted to look as copper. That was found out by chance - I accidentally scratched one of the fins, the paint came off and I saw a silver-like gleam. I suspected it was a fake and scratched the central rod. And found it was made of copper. The Thermaltake's website says the base was made of copper, but nothing was mentioned of the fins.

By the way, the central rod is pressed into the radiator, and for better heat transmission from the rod to the fins some thermal conductive material was used. Such design is a clone of the reference cooler which has a copper core and aluminum fins.

As regards to the base, it is quite big and smooth. No deep traces of polishing is there, and it is too far from the mirror-like perfection.

The cooler has a number of flaws. The most critical is the high noise level. Besides, it is only the core that is made of copper, but the fins are of aluminum. As a result, the cooler showed poor efficiency. And the price of the cooler is twice as much of the aluminum cooler.

Tests

* The fan does change its rotational speed with temperature.
** Noise in the standard mode is impossible!
@ Manufacturer's data
@@ Note that the BOX'ed cooler touches the processor through a layer of foil interleaved, so the processing quality in this case does not play the leading part.

Thermaltake Volcano 7+ made of copper was best at cooling, and Thermaltake Dragon was far worse. At the maximum speed, Volcano 7+ (started with CPUBurn) showed fantastic results, 47 C. This is as high as the aluminum coolers show in the standby mode.

Due to its powerful fan, the best among aluminum coolers was Thermaltake Volcano P4 . But the boxed Sanyo cooler was the best at the 'cooling/noise' ratio. Of note is the Titan TTC-W2T cooler - although it was not able to regulate the fan's rotational speed, showed excellent efficiency and low noise levels (not higher than that of the boxed one). Elan Vital FSNW01-FC is able to change the rotational speed, but that does not help much. I wouldn't recommend buying this cooler because of its poor thermal interface, high noise levels (even at low rpm's) and high price.

Conclusions

If you intend to run your Pentium4 at the rated clock speeds or at increased with the Vcore not exceeding 1.65V, then you can freely install an aluminum radiator cooler.

The threshold value Vcore = 1.65V is taken as a guide. Because any processor can easily withstand a 10% voltage increase.

Of these coolers, the best is the boxed Intel cooler (normally, manufactured by Sanyo). Other cooler may reduce the temperature by a couple of degrees (which may result in increased noise), but that won't produce a positive effect.

But if you are after a heavy overclocking, with the Vcore substantially increased, then a copper radiator cooler is needed. Of all the coolers tested I can recommend only Thermaltake Volcano 7+ whose radiator is able to withstand very heavy processor overheating. Another one, Thermaltake Dragon, because of its high noise levels and aluminum fins can't be recommended for purchasing.

Addendum 1

Addendum 2

Cooler base, factory-made.

After applying grinding powder.
The unwanted concavity in the base center was eliminated.

the base of mirror-like quality

Here comes stage 2: grinding with the GOI polishing paste. Apply some paste on a soft cloth with which you will polish the surface to the mirror-like gleam. The second stage is over as soon as you see your happy smiling face on the radiator base :))).