A Roundup of cheap coolers for powerful processors (SocketA): Q1-2`2003

Author: Aleksandr Mitrofanov
Date: 09/07/2003

Introduction

This roundup is about cooling new AMD processors built on the Barton core. Reminding it to you that this core is a modified Thoroughbred with the L2 cache size increased to 512K. While on the date of release these processors were aimed at the high-end sector, after the recent price reduction they became affordable to the mainstream user as well. The lower-end model - Athlon XP 2500+ - is currently priced at below 100$ (as per PriceWatch.Com).

So today we'll be looking into cheap coolers priced below $13. Mind you, some might think it's a bit too high for a piece of metal with a propeller attached, but keep in mind that Barton processors run a bit hotter than Thoroughbred ones. Besides, prices for AMD processors are going down swiftly, so a good cooler has got to do its job in a "future-proof" way :). In other words, when upgrading to a newer and faster processor we shouldn't worry about replacing the cooler.

Now regarding the measurement procedure. Without any noise measurement instrument on hands, I am going to assess the noise levels and rate it in score points. That is, the higher the score, the higher the score, the quieter a cooler is. Such assessment is quite subjective. I must the author (myself :) is very sensitive to the noise level since the computer is running on a 24/7 basis and must not distract from good night's sleep. Now regarding the temperature. The processor temperatures and fans' rotational speeds were gauged with the Motherboard Monitor v5.2.0.0, and the testing system was assembled in an Inwin S506 housing. If a cooler had a related thermal interface, then the thermal interface was removed and then we measured the noise with the silicon-organic paste KPT-8. The processor was heated up with the BurnK7 software, with Athlon XP with the integrated thermal sensor used as the test processor. That gave a very high precision at measuring the temperature.

By the way, while almost all of the SocketA processors are equipped with the internal thermal interface, this can't be said about their support by motherboards. Only very few allow user keeping track of the core temperature. Recently, it's becoming increasingly popular among manufacturers to implement onboard hardware protection against overheating (normally, at the 100-110C), with the core temperature monitoring completely disabled. As we'll see in this roundup, this function is of paramount significance to the overclocking.

While describing the cooler I'll draw your attention to the fastening convenience. Also don't forget about such factors like fastening reliability, radiator size, material used, exterior, processing quality etc. Most of these are fairly subjective as well (especially, the exterior :), so I'll try to present as more illustrations as possible.

Evercool CUD-725

Upon removing the cooler out of the package (coarse cardboard) I noticed its striking similarity with the Titan TTC-CU5TB cooler. Both coolers are equipped with completely identical radiators with the same fastening scheme.

As you can see, the differences between the coolers are only in the fan - and we'll dwell on it in greater detail.

The fan's dimensions are 70x70x25 mm, with the 4200 rpm rotation speed. The air flow that's generated amounts to 42.08 CFM, with the noise level not exceeding 32 dBA. For comparison: the aluminum fan on Titan TTC-CU5TB offers 3500 rpm rotation speed, produces 28 CFM air flow and creates a 28 dBA noise level.

However, these are just paper data of the manufacturer (which sometimes may not be true to life). Under real conditions, the noise coming from both coolers may be enormous. As you remember, in the roundup on Tital cooler I made a conclusion that the cooler can be recommended for buying only if you have a rpm reducing device. So the same recommendation applies to Evercool CUD-725 as well.

You can reduce the rotation speed absolutely safely - the cooler offers a very good copper radiator, with a large number of thin copper fins and well-polished base.

The processing quality of the base is indeed impressive. Upon removing the protective film (another similarity of Evercool and Titan coolers) in the base, you can watch everything like in a mirror-glass.

Over two dozen copper fins are welded to the base, each having around 1 mm in thickness. To increase the rigidity of the fins, 2 holes are extruded in each of them, with the edges of the holes touching the adjoining fins. As a result, this creates an illusion of two copper tubes.

There is a special metal frame to fasten a fan on the radiator.

To make the cooler look more formidable, the frame is painted to copper color. Actually, I didn't like the fastening of the frame to the radiator - upon some pressure the frame might slip off the fastening ledges.

Here is the interim conclusion: the cooler suits to cooling any SocketA processors (including those overclocked). Among the shortcomings, of note is the high noise level in the nominal mode as well as the unreliable fastening of the frame.

Evercool CUE-610

This cooler caught my sight by its small dimensions and fairly high price. "Means there must be some tricky stuff inside", - I thought and asked the sales assistant to show me the cooler at a closer distance :)

Upon a closer look, I found out the following: the cooler has a copper insert in the base.

The processing quality is very good, although not mirror-like. The base is protected with a special film (not shown on the photo) that guards it against scratches. The good contact between the copper and aluminum parts of the radiator did appeal to me. However, sadly, the thickness of the copper plate is still unknown.

The contact between the CPU core and the copper base proved so good that even after removing the cooler no traces of the thermal paste left.

The fins of the cooler are in fact aluminum needles a few millimeters in thickness.

On top of it there is a low-power 4800 rpm 60 mm fan that produces a 20.26 CFM air flow. As per the manufacturer's data, the noise produced is below 32 dBA, which in my view is quite enormous.

Besides the noise, the cooler has another shortcoming - the fastening clip is not very handy at all. First, to fit the cooler you would need a screwdriver; secondly, the stop for the screwdriver is very small. That is why you've got to be very careful installing the cooler.

As you see, the clip has a hook for one socket tooth only. Anyway, the cooler is not very heavy at all, and the fastening meets necessary requirements.

Igloo 2320Pro

This cooler came to our test lab due to the only reason: it was quite curious to see how a cooler with the full aluminum radiator copes with the job of cooling a Barton processor.

Outwardly, there is nothing special about the cooler: it has a radiator, a fastening plate on which there is a fan. In fact, the exterior produced a good impression: distinct colors and the silvery cover of the radiator.

On the radiator base there is a thermal interface layer applied. Whereas in the previous Glacialtech models that was thermal paste with additions of silver (as per the manufacturer's), this cooler uses ordinary thermal paste which we removed.

The processing quality of the base is frustratingly poor.

To safeguard the thermal interface layer from damage during transportation, the base is covered with a special protective lid made of transparent plastic.

Then, there are over 30 quite high fins on the radiator. On top of the radiator, there is a special lid (shown in blue on the photo) on which a fan is fastened.

By the way, the fan is fastened off the center but is slightly displaced. This is made so that to reduce the so-called "dead spot", i.e. an area where the air circulation is very weak.

The radiator has the following overall dimensions: 70x64x57mm, with 244 grammes in mass. More details about the fan: It has 60x60x20 mm of overall dimensions and 4800 rpm rotation speed. The air flow equals 22.8CFM, with the 35 dBA noise level (as per the manufacturer's data).

Personally, I think the noise is too high and unpleasant.

The fastening of the cooler left a dubious impression: on the fastening clip there is a wide finger stop which allows installing the cooler without a screwdriver. But you will definitely need a screwdriver when removing the cooler - you'd have to hook the clip with a blade.

Besides, the fastening clip bends poorly, which adds issues during the removal/installation.

It's still hard to make a preliminary conclusion - you've got to see the results of tests.

Ajigo MF034-032 (AMD Box)

Officially, this cooler is not a "boxed" version of the AMD cooler. But it fully meets all the necessary specifications and one of the first to be certified by AMD.

Outwardly, the cooler is pretty mean, but the first impression may be misleading. First, the cooler base is made of copper and plated with nickel.

Secondly, to fasten the cooler a clip with 3 hooks and quite handy screwdriver stop are used.

Thirdly, the radiator has a great number of thin fins, thus the total area is increased essentially.

On the radiator, there is a 60 mm fan of 4200 rpm rotation speed. The air flow that's generated amounts to 21.22 CFM, with the noise level being 31.5 dB. The noise level can be regarded as acceptable for comfortable work. But prior to making any conclusions, we should look at the cooling efficiency.

Titan TTC-D5TB(F)CU35

Last year, we reviewed the TTC-D5TB(TC) cooler made by Titan. It appealed to me primarily by its superb exterior and low noise levels (owing to the fan of variable rotation speed). So once I heard of the release of new TTC-D5TB(F)CU35 model, of course I bought that for tests.

At first glance, the cooler makes little difference from the predecessor: the fan is colored black, not silvery. And no sooner had I removed the protective film from the base than

I immediately noticed a copper circle around 3 cm in diameter and 6 mm in thickness (as per the manufacturer's data). It's just this circle that's responsible for taking heat off the processor core and distribute it over the aluminum radiator (whose design remained unchanged) as quickly as possible.

The processing quality of the base is above all praise. What I especially liked is that the copper insert has an ideal contact with the aluminum part.

The radiator has a small bulge in the central area.

Another difference is that the fastening clip has a hook for all the three socket teeth (not for one as is the case with TTC-D5TB(TC).

These don't matter much though, since the cooler mass is not great on the whole.

On the cooler, there is a 80 mm fan of 2800 rpm rotation speed. The noise level that's generated does not exceed 32 dBa, with the air flow being 37.57 CFM.

Unfortunately, the fan does not offer a thermal sensor and can't change the rotation speed depending on the processor temperature.

Thermaltake Volcano 8(SE)

This cooler is aimed mostly at assembly operators, rather than overclockers.

At the same time, the cooler has a radiator with a copper plate in the base. This allowed the manufacturer to claim of compatibility of the cooler to any SocketA processor up to Athlon XP2800+ inclusive.

The processing quality of the base is middling: grinding traces are visible. On the base, there is a layer of Bergquist 225U thermal interface applied (no longer visible on the photo), which allows assembly operators doing without thermal paste. But to overclockers I would advise to remove this substance and use regular KPT-8 thermal paste.

As you can see, the copper plate is quite thick, with the aluminum fins (about 20) having large surface area.

A good radiator is merely half success - a powerful fan is a must. Therefore, Thermaltake installed a 60 mm fan of 4800 rpm on the Volcano 8. As per the manufacturer's data, the fan provides a 20.2 CFM air flow with the noise level being 35 dBa.

Volcano8 has another advantage - the fastening clip has a hook for all the socket teeth, which provides a reliable attachment.

Titan TTC-CU6TB

This is another cooler made by Titan whose exterior casts serious doubts as to the cooling efficiency.

Anyway, there is a copper plate welded to the radiator base (it's a pity the thickness is unknown). The contact with the aluminum part of the radiator seems to be good enough, and the polishing quality deserves best praise (ideally smooth, albeit not mirror-like).

Once I removed the cooler out of the box, I noticed that the base is protected with a special film. Actually, all Titan's products (and Evercool's as well) are equipped with such protection.

At that, all the pros of the cooler are over - then there go only cons. First, the fastening clip is completely uncomfortable, and the cooler installation runs the risk of damaging it (because the screwdriver is going to slip off).

More problems are caused by the process of the cooler removal - at that, I even had to take the board completely out of the housing.

The other shortcoming is about the fan. By its characteristics, it appears to be fine: the rotational speed is 4800 rpm, with the air flow being 20.26 CFM. However, in practical tests it demonstrated quite a high noise level (higher than the declared 32 dBA) and poor efficiency. In fact, the poor efficiency of the cooler may be caused by the faulty design of the radiator.

In any case, this cooler can hardly be recommended for cooling the processor. I wouldn't recommend it even for cooling the chipset (e.g. nForce II SPP), since Evercool CUE-610 would do a much better job of that.

Tests

Before testing the coolers I summed up all their specifications into a table (the higher the score, the better).

For tests, we took a Barton processor. The clock speed was set to 2000 MHz (12x166) under normal voltage. Let's see how these coolers cope with such a load.

So, at efficiency the first place was won by the two twins: Evercool CUD-725 and Titan TTC-CU5TB. But their noise levels leave much to be desired - for comfortable operation, one has to use a speed-reduction device. The second place was taken by Thermaltake's Volcano 7. It has a 80 mm fan of variable rotation speed (depends on the processor temperature), however, its noise level is a bit too high. For comfortable operation, I would recommend having the fan replaced with a less noisy one.

By the way, note that currently the Thermaltake Volcano 7 is priced below $13 and thus falls within a low-end group of coolers.

As regards those users for whom the noise levels are of paramount importance, then to my regret I cannot recommend any of the coolers presented. In this price category, there are no truly noiseless coolers. However, there are two that stand out: Titan TTC-D5TB(F)CU35 and Titan TTC-D5TB(TC). The former demonstrates a bit better performance, but makes more noise. The latter runs very quietly (owing to the variable rotation speed), but has a fully aluminum radiator.

As an experiment, we can cross-breed these two coolers, but the cost of the result will exceed $20 :)

Besides, Ajigo cooler demonstrated acoustic performance, but I am not sure it can be made available in the retail over Russia.

As for Thermaltake Volcano 8 and Evercool CUE-610 coolers, they appeared to be firm "middlers" which could be safely used in mainstream systems without overclocking in mind.

And finally, the Titan TTC-CU6TB and Igloo 2320Pro coolers cannot be recommended for buying at all.

So, the totals have been summed up. Now turn to a very interesting data on the difference in temperature readings taken from the external thermal sensor as opposed to the internal. The amount of the difference is not constant and strongly depends on the cooler in use.

For instance, for the Ajigo MF034-032 cooler it is as high as 2 C, while for the Titan TTC-D5TB(F)CU35 it is over 10 C. What is the cause of that?

The thing is that the external thermal sensor is positioned practically on the board's surface (e.g. Soltek KT400A). This means its readings will be affected by the temperature of the part positioned near the processor socket. Of course, if the air flow generated by the cooler cools down the board, then the external thermal sensor will also show a lower value. The second factor that affects the difference between the readings of the internal and external thermal sensor is the material's thermal resistance from which the radiator is made. Simply spoken, at which rate the radiator transmits heat from the processor core. Take a look at the Titan TTC-D5TB(TC) cooler - its radiator is made of aluminum, and the fan is sizeable (80 mm). As a result, the difference in the readings of thermal sensors amounted to 13 C.

I calculated the difference in readings between the internal and external sensors for all the coolers, and for better clarity presented its information graphically:

Besides, the readings of the thermal sensors (internal and external) were much influenced by the motherboard itself and its BIOS. First, the thermal sensor may be positioned on the board's surface while being near the processor. Example:

preliminary and final revision of the Epox 8RDA3+ motherboard.

Secondly, the board's BIOS may add serious corrections tot he readings of the sensors. It happens very often that upon flashing the BIOS with a new version the readings of the temperature sensors are corrected.

Somehow or other, but the main conclusion is this: for hardcore overclocking, the readings of external thermal sensor are absolutely useless since they don't reflect the factual processor core temperatures. It's absolutely real when the user replaces one cooler with another, with the external thermal sensor understating the temperature. In fact, the core temperature remained the same (or even rose). This results in overheating and deterioration of the overclocking results.

As for the ordinary users (indifferent to overclocking), they would be quite content with the readings that come from the external thermal sensor. If it indicates a stable temperature, no more than 60-70 C (under load), then the cooler is installed in a correct manner, and the fan functions properly.