Green is a hot adjective. In fact, it was one of the fastest growing buzzwords in 2008. Across every industry, green is the new black, being slapped on products that destroy the earth only slightly less than other products.
There are truly green products though, and anything that can cut down on electricity usage in a significant way deserves to be called green. In addition to a spiffy green logo, AMD has been putting out low-TDP processors that provide lower power and cooling requirements with all the horsepower of their regular chips since 2006.
What is TDP, and why does it matter?
Rather than gloss over it, Icrontic has given this topic the proper treatment in another article. In summary, lower TDP means a chip is able to operate reliably below the voltage it was engineered to operate at. This lower voltage results in cost savings due to lower energy usage, lower heat emissions, less noise from the cooling system, and better overclocking due to stability at a lower voltage and thermal envelope.
Low-TDP chips are a boon to businesses, those with environmental concerns, silent PC enthusiasts, home theater PC builders, and gamers alike.
History of AMD’s low-TDP ‘no compromise’ parts
Oftentimes, low-TDP chips are stripped down versions of existing chips. A dual core chip becomes a single-core low-TDP model, or a chip is engineered with less cache and slower clockspeed. AMD has, however, produced no-compromise low-TDP chips since the introduction of the Athlon 64 X2 3800+ EE SFF chip in the fall of 2006. This CPU was a 35W chip that provided the same performance as its 90w brethren. Hewlett-Packard essentially monopolized the supply of the chip for almost a whole year, allowing them to offer their Slimline series desktops with dual core chips at a time when dual core was still a pretty high-end feature.
The Brisbane core made its debut in late 2006 in 65W Athlon 64 X2 parts, and found its way to a low-TDP designation by mid-2007. Released initially as the BE-2xxx models, they were subsequently renamed to fit the 4-letter designation AMD adopted in April 2008. A total of nine 45W dual core processors were launched, the last of which, the 2.6GHz Athlon X2 5050e, was released in October 2008.
The new coolness
Today, AMD is releasing no less than six new 45W processors under the Athlon II X2 nameplate. This batch of low-power processors is the first to be introduced in almost a year, after more than 3 months of zero supply for 45W parts via distributors.
Like the past versions of greener CPUs, AMD isn’t reinventing the wheel for these chips. They’re taking existing cores, eliminating the L3 cache, and finding the cores that operate at lower voltages. The Athlon II X2 is codenamed Regor, which is a cut-down version of the Deneb core found in the Phenom II X4. The Athlon II X4 (Propus) and Athlon II X3 (Rana) are also Deneb derivatives. Like the Brisbane core before it, testing after manufacture discovers a number of the processors are stable at lower voltages, which allows them to be binned as ‘efficient’ chips.
What differentiates the Athlon II parts from the Phenom II parts is primarily their lack of L3 cache. This reduces the size of the processor by reducing transistor count; having fewer circuits to power means the chip requires less voltage to operate. The 235e and 240e compensate somewhat by adopting the L2 cache that would’ve been allocated to sister cores in an X4 model. The 400e, 405e, 600e, and 605e get the benefit of additional cores but must cope with only 512K of L2 cache per core. Our review will focus on the dual core 45w parts.
Evaluation
The Athlon II X2 240e is functionally identical to and will provide the same performance as the original Athlon II X2 240. There are a few distinct types of users who are going to be interested in this low-power chip, and we decided to use this review as an opportunity to view things from those angles.
- Owners of existing low-power parts who could use a bit more oopmh will ask themselves “Is this going to provide a substantial performance increase from my older low-TDP CPU?”
- Owners of AM2+ motherboards who would see benefit from heat/noise/electrical cost reduction will wonder “What’s the thermal and acoustic performance like?”
- People looking to build new systems who would see benefit from heat/noise/electrical cost reduction will benefit from our upcoming look at performance on the AM3 platform.
Methodology
In order to find out how the new CPUs perform for people with existing AM2+ motherboards, this first round of testing will focus on performance on the AM2+ platform. All of the parts used are pretty typical and midrange, to give a very average test case. We’ll be looking into performance on the AM3 platform in an upcoming review, for those looking to deploy brand new systems.
Hardware used for test:
- Gigabyte GA-MA74GM-S2 rev 2.0
- Samsung 2 x 1GB DDR2 800MHz
- MSI Radeon 4350 512MB (R4350-D512Hv2)
- Western Digital 320GB SATA 16MB Cache 7200RPM (WD3200AAKS)
- Toshiba Samsung DVD-ROM TS-H353
- Antec Basiq 500W ATX12V PSU
The software and versions used:
- PCMark Vantage – Version 1.00
- 3DMark Vantage – Version 1.01
- Cinebench 32-bit – Release 10
- X264 Benchmark – Version 2.0 (v0.59.819M)
- Sisoft Sandra 2009 – SP3 (build 15.99)
The CPUs were given a dab of Arctic Silver Céramique thermal compound, and hooked up to a standard retail box AMD heatpipe cooler. Cool ‘n’ Quiet was disabled in order to maintain the highest clockspeed possible. Because we wanted to isolate the CPU performance and were not trying to ascertain gaming performance, a very basic GPU was used in the test bed. The motherboard was run on the bench without any chassis. The rest of the components are listed below.
Windows 7 Ultimate was installed fresh, along with the benchmarking and diagnostic apps. Windows Update was disabled, but all other Windows services were left untouched.
Free system memory was checked before each test was run off of a fresh reboot to ensure the testing environments were identical. Speedfan was used to keep logs of CPU and System temperatures, along with the fan speed.
SiSoft Sandra 2009
In order to see how the CPU stacks up in various synthetic measures, we ran the Arithmetic, Multimedia, and Cryptographic tests, along with the Memory Bandwidth and Memory Latency tests.
3DMark Vantage CPU Score
This test attempts to render various scenes and textures with only the CPU. It’s not a reflection of real-world gaming performance so much as it is a raw processor performance benchmark.
PCMark Vantage Sub scores
Rather than look at the overall score, we dropped the gaming and hard drive benchmarks and chose to focus on the individual category sub scores, which run basic everyday tasks to see how the processors stack up. These are much more ‘real-world’ than some other benchmarks, because they don’t always scale well to multi-cored CPUs, as is the case with many applications.
Cinebench 32-bit
3D rendering is a great way to show what a CPU has up its sleeve. Cinebench tests ray-tracing rendering prowess on one core, then all cores, and finishes with an OpenGL test using the system’s GPU. The OpenGL test won’t be very useful for these benchmarks due to the performance limitations of the Radeon 4350 video card that was part of the test bed.
x264
The proliferation of amateur video has led to video encoding being much more commonplace than it once was. The x264 test evaluates a processor’s power by seeing how quickly it can take a 720p video clip and encode it to MPEG4, one of the most common high-definition video codecs.
The x264 benchmark makes two passes at encoding to see how a CPU handles the job.
Thermal and acoustic Performance
We’re not going to get too in-depth here, but we wanted to make sure we at least touched on heat levels and noise. Based on logs kept during testing, these charts show the highest temperature achieved by the core and the highest fan speed achieved. This will be an incomplete picture, for several reasons, but here are the results.
A few things to note about these numbers:
First of all, it doesn’t reflect that in the real world, these CPUs are designed to clock down and run at reduced voltages when necessary. As stated above, Cool ‘n’ Quiet was disabled during testing, so even though we’re comparing a 45W processor to a 95W processor, that doesn’t mean that at 100% CPU utilization, it’ll be 50% less hot.
Additionally, this information won’t reflect the average temperature or fan speed during testing. While the upper limit on these temperatures wasn’t too far apart, the duration spent at those temperatures and at what fan speed could differ greatly.
Analyzing this would’ve taken much longer than we had before press time, but we look forward to providing a more complete picture in the future.
Final thoughts
This CPU is not going to be for everyone. As we learned in our recent Athlon II X4 620 review, you can get a quad core for around the $100 mark. With the Athlon II 240e debuting at $77, its price is within striking distance of the 620. Nonetheless, the 240e and its siblings definitely have their niche. Typical business machines don’t need double-barreled performance, let alone four cores, and when dealing with corporate quantities of computers, the energy reduction from a 20W TDP difference can stack up to some real savings. Home theater PC and silent computing enthusiasts are a captive market for a processor that can lower noise noticeably. And they’re the only game in town for those looking to reduce their carbon footprint while still being able to get their work done.
Price isn’t always the only consideration, and in any situation where efficiency is a premium unto itself, these e-series chips will find a natural home.