by Ilya Gavrichenkov
08/02/2006 | 04:17 PM
The first impressions left by the new Core 2 Duo processors from Intel that arrived to replace not the best Pentium 4/Pentium D processor families turned out to be highly positive. Since the Intel processors with Core microarchitecture have been officially launched already, we had a lot of opportunities to check out their strengths and attractive features. Core 2 Duo CPUs demonstrated unprecedented performance, relatively low heat dissipation and excellent overclocking potential. Everything indicated that a new favorite appeared in the processor market. It is true, all objective data proves that Athlon 64/Athlon 64 X2 processors that used to offer the best combination of consumer-friendly features cannot be a serious competitor to the new Intel CPUs on Core microarchitecture anymore. However, the actual situation is not quite like that.
AMD doesn’t really feel defeated at all. On the contrary, the company is revising its strategy so that its CPUs didn’t lose any of their attractive features against the competitor’s background, but gained some new ones instead. Thus, AMD Athlon 64 and Athlon 64 X2 processors retained their attractiveness for the customers thanks to the significant price reduction that touched upon all AMD processor line-ups and took place together with the Intel Core 2 Duo launch. Although they cannot be positioned as solutions for the high-end computer systems any more, they still remain a great choice for mainstream platforms, which we have already discussed in our article called Contemporary Dual-Core Desktop Processors Shootout.
However, AMD processors have suddenly acquired one pretty serious drawback against the background of Intel’s new CPUs. AMD Athlon 64/Athlon 64 X2 used to be the most economical solutions compared with CPUs on NetBurst architecture. Now that Core 2 Duo appeared, AMD processors turned out pretty power-hungry all of a sudden. Of course, all the obvious consequences followed: such as the need for noisy, technologically complex and expensive cooling solutions and high-capacity power supply units.
However, we should give due credit to AMD: their engineers managed to save the situation. Namely, they will very soon move the special niche economical processor models with the 65W and 35W heat dissipation into the fully-fledged mass market. Although AMD’s Energy Efficient processors are still not available in retail, we were lucky to get a few samples for our tests, and today we are going to tell you we can find out about them. Hopefully, Athlon 64 X2 processors with lower heat dissipation level will be able to successfully compete with the youngest Core 2 Duo models not only from the price-to-performance prospective, but also from the today’s popular “performance-per-watt” prospective.
In the beginning AMD was going to launch the Energy Efficient processor family including single-core and dual-core models for the quiet and economical systems only. These CPUs would be typically installed into multimedia home systems assembled with AMD Live! concept in mind that implies the acceptable noise level of 30dBA.
AMD announced energy efficient CPUs for Socket AM2 systems back in mid May. At that time they launched two processor families: with 35W and 65W maximum heat dissipation. This classification is still valid. Today the first group of the CPUs includes relatively powerful dual-core processors working at up to 2.4GHz frequencies and rated as 3800+, 4200+ and 4600+. These Energy Efficient processors dissipate about the same amount of heat as the Sempron CPUs, which makes them 37% more efficient from the “performance-per-watt” prospective than the regular Athlon 64 X2. The second group of processors is positioned as less high-performance but more economical solutions for Small Form-Factor systems and includes a single dual-core processor with 3800+ performance rating and a few less interesting for us single-core Athlon 64 and Sempron models.
Note that there is nothing extraordinary that Energy Efficient AMD processors appeared in the market. It is known that there is a linear dependence of the CPU heat dissipation on processor clock speed and squared dependence on CPU Vcore. These ratios help the processor developer to create more economical choices. The CPUs with 65W maximum heat dissipation work at slightly lower clock speeds and support lover Vcore of 1.2V-1.25V. This is exactly how they manage to reduce the thermal power of the dual-core processors from 85W to 65W. As for the dual-core Athlon 64 X2 3800+ with the maximum heat dissipation of 35W, it is the 1.075V Vcore and only 2.0GHz clock rate. In other words, Energy Efficient AMD processors are nothing completely new. The manufacturer simply sorts out at the packaging stage those semiconductor dies that can work stably at lower Vcore. After that they simply lower their core voltage and limit the maximum clock frequencies and voila! – you get the desired wattage.
Therefore, it means that you shouldn’t try to find any specific unique features in the Energy Efficient processors. From the microarchitecture and performance standpoints they are exactly the same as the regular processors. Moreover, they are manufactured with the same production technology that is used for standard CPUs. In fact, AMD Energy Efficient processors appeared solely thanks to the improved 90nm process used on Fab.30, which accommodates AMD’s major production capacities.
Keeping in mind that the maximum heat dissipation of the new Intel Core 2 Duo processors is 65W, AMD decided to get more aggressive about pushing its Energy Efficient solutions into the market. Of course, this is a very good marketing move especially since there are no technological obstacles that could prevent it from happening. With these actions in mind that company is already rearranging the major Athlon 64 X2 desktop processor family, so that it could logically contain both: the traditional Athlon 64 X2 alongside with their analogues with 65W TDP. These solutions may be regarded as Core 2 Duo alternatives not only from the price-to-performance prospective, but also from the “performance-per-watt” standpoint. Although AMD has set the prices on the new Energy Efficient solutions slightly higher, the users building quiet computer systems will hardly be scared away by that.
We are also getting ready for the arrival of the Energy Efficient processors with 65W and 35W maximum heat dissipation. First of all, let’s check out the marking differences. The processor TDP is indicated by the third letter in the OPN line and the third symbol in the processor marking respectively:
The formal characteristics of the dual-core Energy Efficient CPUs are the same as those of “regular” processors except the core voltage and heat dissipation section:
We got our hands on two dual-core Energy Efficient AMD processors. One of them was Athlon 64 X2 4600+ with 65W maximum heat dissipation:
Another one was Athlon 64 X2 3800+ with 35W maximum heat dissipation:
As we have expected, the CPUs differ from their regular brothers by the third symbol in the marking that indicates their type. The diagnostic CPU-Z utility didn’t reveal any other specific peculiarities of these particular processors, besides the processor Vcore:
In other words, you cannot really tell the Energy Efficient processors without measuring their electrical and thermal parameters, because they are absolutely the same as the regular CPUs. Just like the common processor models. Energy Efficient Athlon 64 X2 4600+ works at 2.4GHz clock frequency and features 512KB on-die L2 cache, and Athlon 64 X2 3800+ works at 2.0GHz nominal speed and features the same L2 cache.
I have to say that overclockers pin special hopes upon AMD Energy Efficient processors, as they expect these CPUs to boast a much higher overclocking potential. It is pretty hard to determine what these expectations are based on, especially since Energy Efficient processors feature the same microarchitecture as their “regular” counterparts. And the stable operation at lower nominal clock frequencies and with lower Vcore doesn’t at all mean that it will work at sky-high frequencies if you increase its core voltage…
However, we couldn’t help checking out the overclocking potential of the processors that we got at our disposal. Athlon 64 X2 4600+ with the maximum heat dissipation of 65W was the first one we installed into our testbed.
The testbed was built using ASUS M2N32-SLI Deluxe mainboard based on Nvidia nForce 590 SLI chipset, Corsair Twin2X1024-8500 memory and PowerColor X1900 XTX 512MB graphics card. The processor was cooled with Zalman CNPS9500 AM2 air-cooler. Since the maximum clock frequency multiplier of the Energy Efficient processors (just like the “normal” ones) is locked, these CPUs can only be overclocked by raising the clock generator frequency. In this case we had to reduce the multiplier for the memory frequency and HyperTransport bus connecting the CPU with the SPP in order to ensure that all other subsystems of the test platform work stably.
Without increasing the core voltage, i.e. with 1.25V of power, Athlon 64 X2 4600+ working at 2.4GHz nominal speed could only overclock to 2.6GHz (and work stably at that frequency). The clock generator frequency in this experiment reached 221MHz.
The second overclocking experiment was conducted with the processor core voltage increased to the level of “regular” – not Energy Efficient – CPUs, that is to 1.35V. In this case we managed to raise the clock generator frequency to 232MHz without losing the system stability. The CPU worked at 2.78GHz in this case.
The third experiment was performed with the processor Vcore raised to 1.5V. This core voltage is 20% higher than the nominal value, and normally you should be concerned with arranging special cooling for the processor. However in this case it is pretty safe, because the Energy Efficient CPUs are manufactured using the same semiconductor dies as the regular processors, so the Vcore increase to 1.35V can be performed safely, as the potential of the die “allows” it initially. In other words, it makes sense to increase the core voltage basing on the nominal Vcore settings for the “regular” processors from the same family, rather than on the nominal Vcore set for the given CPU model.
We managed to get our processor to work at 2.84GHz core frequency while the clock generator frequency in this case equaled 237MHz. all further overclocking attempts to hit higher speeds resulted in instability, so we would consider the last result to be the maximum.
This way, Energy Efficient Athlon 64 X2 4600+ processor with the maximum heat dissipation of 65W could overclock only to 2.84GHz, i.e. its frequency potential didn’t surprise us. The regular CPUs of the same type with the 85W TDP could most probably hit the same mark during overclocking. In other words, our very first experiment indicated clearly that there will hardly be anything special about the overclocking potential of the Energy Efficient processors. However, it is definitely too early to make any conclusions. We still have one more Energy Efficient processor: Athlon 64 X2 3800+ with the maximum wattage of 35W and 2.0GHz nominal clock speed.
The nominal Vcore of this CPU model is 1.075V, so we didn’t really expect much without increasing this parameter. And our expectations were absolutely true: by simply raising the clock generator frequency we could only hit the 215MHz mark, which results into 8% overclocking to 2.15GHz.
If the processor Vcore were increased to the “normal” level of 1.35V, then it immediately acquired a better overclocking potential. However, the maximum result we managed to achieve in these conditions was also not that impressive. The CPU could work at 2.3GHz maximum without losing its stability.
Further Vcore increase to 1.5V allowed us to raise the CPU core frequency even higher. However, once again we have to admit that the overclocking potential of our Athlon 64 X2 3800+ Energy Efficient processor appeared far below our expectations. The maximum frequency we managed to get it working stably at – was only 2.55GHz.
And this is the result we managed to obtain with 40% core voltage! Despite all our efforts, the clock frequency got only 27.5% higher.
This relatively scarce acceleration together with the results of our previous experiment destroys the opinion that Energy Efficient processors would be a great choice for overclockers. Both experiments described above show just the opposite: Energy Efficient processors overclocker not any better than their “standard” counterparts. In other words, it doesn’t make much sense to hunt for Energy Efficient CPUs if you intend to overclock them. Hitting extreme frequencies is not among their strengths, their advantage is of totally different sort: they can work at relatively high frequencies with lower core voltage, which eventually reduces their heat dissipation and power consumption.
The thing is that the differences between the families of processors with different thermal power lie in their maximum clock frequencies and Vcore. Moreover, if you recall the well-known linear dependence between the maximum CPU frequency and Vcore, everything will be absolutely clear right away. It is exactly the different Vcore that determines the maximum frequencies. Therefore, you shouldn’t believe that since Energy Efficient processors have lower nominal voltage than the “standard” ones, they will overclock better thanks to bigger range for Vcore increase. In reality if you raise the Vcore of the Energy Efficient processor to the level of the nominal Vcore of a regular one, you will simply be able to catch up with the frequency of the regular processor, but not overcome it.
This is where AMD processors are very much different from Intel ones. Intel CPUs comply with the rule that the processor with lower nominal Vcore overclocks better than the processor with higher nominal Vcore. And it is actually very easy to understand why so. By launching the same CPU model with different core voltages Intel pushes the core frequency parameters to the desired frequency by increasing the Vcore. So it is absolutely clear in this case, that if the CPU hits the desired frequency at a lower Vcore, then it will definitely overclock better. AMD uses the opposite approach: the Vcore gets lower together with the clock frequency just to reduce the heat dissipation. In this case, there is no way the frequency potential gets improved.
We decided to carry out a few practical tests to estimate how economical the new AMD Energy Efficient processors are. Therefore we compared the current power level in platforms equipped with different processors. We measured the currents going through the processor voltage regulator circuitry (this way we could check the processor power consumption without the processor voltage regulator efficiency) as well as through the entire system. As a result, we will be talking about two parameters: CPU power consumption and platform power consumption.
For our tests we used the following processors: regular AMD Athlon 64 X2 4600+ with 85W TDP, Energy Efficient AMD Athlon 64 X2 4600+ with 65W TDP and Energy Efficient AMD Athlon 64 X2 3800+ with 35W TDP. Our test platform was built using ASUS M2N32-SLI Deluxe mainboard based on Nvidia nForce 590 SLI chipset, 1GB DDR2-800 SDRAM, PowerColor X1900 XTX 512MB graphics card and Western Digital Raptor WR740GD HDD. The processor was cooled down by Zalman CNPS9500 AM2 air cooler.
To compare the obtained results against the power consumption of the competitor’s solutions, we measured the same parameters in the platform built with the new Intel processor on Core microarchitecture. The alternative competitor solution was built around Intel Core 2 Duo E6300 processor, which performance is almost the same as that of AMD Athlon 64 X2 4600+, and ASUS P5W DH Deluxe mainboard on Intel 975X chipset. The system was also equipped with the same graphics card, memory and hard disk drive. The CPU was cooled down with a similar air cooling solution from Zalman – CNPS9500 LED.
We measured the systems power consumption in different modes: in idle mode, during WMV video-movie playback from the hard disk drive, in a popular 3D benchmark – 3DMark06 and under maximum CPU and memory workload created by a special S&M utility.
So, first of all we measured the power consumption in idle mode with Cool’n’Quiet and Enhanced Intel SpeedStep activated.
In idle mode and with all CPU power-saving technologies enabled, AMD processors consume less power than the youngest Intel Core 2 Duo processor model. It may be determined by the fact that Cool’n’Quiet technology allows Athlon 64 X2 to drop the clock speed to 1.0GHz in idle mode, while Enhanced Intel SpeedStep lets Core 2 Duo drop its clock speed to only 1.6GHz.
If we disable the Cool’n’Quiet and EIST technologies, the situation gets slightly different:
The power consumption of AMD processors with 85W TDP gets much higher and they start eating up more power than their Intel rival, which power consumption increases just a little bit thanks to the Enhanced Halt State (C1E). As you remember from our previous articles, C1E technology just like EIST allows the CPU to drop its clock speed on Halt command from the OS. That is why Intel processor consumes so little power in this test.
Now let’s see what is going to happen if we load the CPU with work. First we will measure the power consumption during WMV-video playback in Windows Media Player 10.
Video playback is hardly a hard task for the system CPU. Therefore, the picture we observe here is very similar to what we saw in idle mode. Energy Efficient AMD processors turn out more efficient than Intel Core 2 Duo, while the “standard” Athlon 64 X2 loses to its brothers as well as to the competitor.
If we load the testbeds with harder tasks, such as gaming applications, for instance, the situation will change:
Energy Efficient Athlon 64 X2 4600+ with the 65W TDP starts to fall behind the Intel Core 2 Duo E6300 processor from the “performance-per-watt” standpoint. Only the 35W Athlon 64 X2 3800+ manages to prove its name (in terms of power consumption). However, its performance is about 10-12% lower than that of the Core 2 Duo E6300 and Athlon 64 X2 4600+.
In conclusion let’s take a look at the power consumption measurements when the CPU is loaded heavily by the S&M utility.
What you see on the chart is a complete failure of the K8 architecture. When the CPU is fully loaded, even the most economical AMD processors consume more power (and hence dissipate more heat) than the Intel Core based CPU with the same performance level.
However, for the dedicated AMD fans I would like to mention a few facts that may change the attitude to the last chart. The thing is that S&M, just like many other tools creating ultimate processor workloads, use special floating-point operations. And it was fine for CPUs on K8 or NetBurst microarchitecture. However, Core 2 Duo processors do not get loaded to the full extent with these utilities. We had to go through a number of different burn-programs to realize that there are a few old ones that do the job much better than S&M, prime95 and others.
For example, when we resorted to BurnK6 tool that we used to use to heat up AMD K6 processors back in the days, we managed to get much higher power consumption numbers for Core 2 Duo E6300: 55W for the CPU and 229W for the platform. In other words, Intel Core based processors hit the maximum power consumption and heat dissipation in absolutely different type of tasks than their competitors and predecessors.
In conclusion we would like to provide the numbers we got when the memory controller was loaded to 100%:
The picture is very similar to what we have just seen in the previous test. Core 2 Duo processor again proves that Intel Core microarchitecture is very economical. And as we see, the same result is obtained for the CPU and the entire platform, which means Athlon 64 X2 processors cannot blame their specific design, namely the integrated memory controller, for this failure.
Summing up I would like to admit that being economical is the major advantage of AMD Athlon 64 X2 Energy Efficient processors. They really consume less power and dissipate less heat than their standard counterparts. The CPUs with 65W TDP consume 25-30% less power under workload, while the CPUs with 35W TDP allow you to save about 55% of power. Of course, these features of the Energy Efficient make them ideal for small and quiet systems such as getting more and more popular home multimedia centers, entertainment centers and small form-factor PCs.
Unfortunately, Energy Efficient cannot boast anything else besides the lowered thermal and electrical characteristics. They do not differ in architecture from the “standard” Athlon 64 and hence perform absolutely identically. As for their overclocking potential, which might be of specific interest to hardware enthusiasts, we have to upset you. Despite the expectations, Energy Efficient AMD CPUs cannot boast higher overclocking potential. They may even overclock worse than their “standard” analogues. Although their lowered power consumption allows you to raise the core voltage much higher, it doesn’t help to hit the extreme clock speeds.
Closing up our review I would like to add that from the heat dissipation and power consumption standpoints, the AMD Energy Efficient processors can easily compete with the new Intel Core 2 Duo CPUs based on Core microarchitecture, which offers extremely efficient performance-per-watt ratio. 65W Athlon 64 X2 consume less power under low workload than the Core 2 Duo units with the same performance level. As for AMD processors with 35W TDP, they turn out absolutely beyond any competition from the electrical and thermal prospective, although you should always keep in mind that at the same time they are not so fast as the competitor.
By the way, AMD processors with 35W TDP really impressed us with their thermal mode. The screenshot below was taken during our test session. The numbers speak for themselves:
All in all we are ready to voice our verdict: Energy Efficient AMD processors will be interesting only for those platforms where low power consumption and heat dissipation are critical. In all other cases they may not be the best choice keeping in mind their higher price.
