Energy Efficient processors from AMD (they consume 65 Watt maximum, while regular Athlon 64 X2 processors consume 89 Watt) appeared rather long ago, but... we were too busy to review them. :) Today we present an article about three EE processors: 4400+, 4600+, and 4800+. Besides, two of them are manufactured by the 65-nm process technology (4400+/4800+). Theoretically, it should have a positive effect on their power consumption. Our inquisitive readers will finally learn about performance of the new 4400+ model (modern Athlon 64 X2 with this model number differs in technical characteristics from the old modification).
There is a little intrigue in this review: the 65-nm Athlon 64 X2 processors come with L2 cache with worse latencies. Besides, processors with non-integer multipliers (the new 4400+ - 11.5, the new 4800+ - 12.5) do not work with memory at full frequency owing to architectural limitations of the built-in memory controller. So, the real frequency of DDR2-800 memory with the 4400+ is just 767 MHz, with the 4800+ - 714. Thus, there are a number of preconditions for lower performance of the new modifications.
Representatives of the other platform were chosen in a traditional way: the total average score of one of them must be lower than the worst result of our processors under review, and the average score of the second must be higher than the biggest score of our processors under review. Thus, you can see how these AMD processors fare - their performance is in between the competing processors we have chosen.
Hardware and Software
* - "2 x ..." means per core
Essential foreword to charts
Our test method has two peculiarities of data representation: (1) all data types are reduced to one - integer relative score (performance of a given processor relative to that of Intel Core 2 Duo E4300, given its performance is 100 points), and (2) detailed results are published in this Microsoft Excel table, while the article contains only summary charts by benchmark classes. We will nevertheless focus your attention on detailed results, when needed.
3D Modeling and Rendering
The first results are shocking: the new 4400+ is outperformed by the old, the latter is even faster than the new 4800+! Only the old 90-nm 4600+ EE is doing fine: quite a decent result that goes well with the results demonstrated by the 5000+. But it's based on the old core with a faster L2 cache. It also has an integer multiplier, so its memory operates at standard 800 MHz.
This situation is better, but it's still far from being ideal: the 4800+ EE is not outperformed by the old 4400+ anymore, but it's still slower than the 4600+ EE. The new 4400+ is again outperformed by the old modification.
Digital Photo Processing
Judging by the results, a core clock rate is more important for Photoshop than memory frequency and cache size: the new 4400+ finally manages to outscore the old modification at least by one point. The 4800+ also manages to catch up with the 4600+.
This situation is as bad as on the first diagram. There are many reasons for that, they all play against the 65-nm core: the compiler is sensitive to cache size, and to its speed, and to memory bandwidth.
It's another test critical to memory bandwidth and cache size, so the results are predictable. I don't want to generalize, but two identical columns with 92 points suggest sad ideas about "model number inflation": the old 4400 processors perform on a par with the new 4800...
Everything is fine in RightMark for an evident reason: performance grows almost proportionally to the clock rate (within the same architecture); this test practically ignores cache size and memory bandwidth.
I think you already predicted sad archiving results: this class of tasks is critical to cache size and performance as well as to memory bandwidth.
It's another application, which seems to be sensitive only to the core clock. The result is predictable: indeed, the clock rate of the new processors is a tad higher than in the old modifications.
It's an old test. It shows well where the 65-nm processors from AMD will prosper: in old applications, which do not require much from memory bandwidth and cache.
Video codecs are also quite loyal to the "core clock is above all" concept.
...Games strongly object to this concept. It was evident even in the comparison Socket AM2 versus Socket 939: faster memory helped the new socket. And now it plays a nasty joke on the new processors from AMD: games prefer honest DDR2-800.
On the whole, the situation with performance is quite dismal. A new processor with model number 4800+ being outperformed by a processor with model number 4600+ is the first sign of the most sceptic forecasts: model numbers have nothing to do with performance. It would have been fine if the advantage of the 4800+ over 4600+ was just not proportional to the difference between these model numbers. But the 4800+ being slower?! AMD have introduced a new naming concept for a reason: the old one has discredited itself.
Estimated power consumption
At least here everything is fine: results of the old 90-nm Athlon 64 X2 4600+ EE are quite good even compared to competing Intel processors, and the 65-nm 4400+ EE and 4800+ EE look splendid: power consumption in the idle mode lets them compete even with Core 2 Duo and Pentium E. However, Pentium E2160 is still victorious under 100% load. At the same time, Athlon 64 X2 4400+ EE and Core 2 Duo E6420 both demonstrate 39 Watt, a respectable result as well: AMD have made great progress in designing "cold" CPUs.
On the whole, Energy Efficient processors from AMD are indeed energy-efficient, especially the 65-nm models. They don't reach 65 W even under 100% load, and offer excellent power consumption in the idle mode. Almost like Core 2 Duo. ;) Athlon 64 X2 4400+ EE consumes even less power.
However, their performance results look pitiful on the background of their power consumption success. The new 65-nm 4400+ is generally slower than its old modification, despite its higher clock rate. The 65-nm 4800+ is a waste of money. Only the 4600+ fares well... However, it's a 90-nm processor, even though it bears the Energy Efficient title. By the way, it's not the best choice as far as power consumption is concerned (AMD do not offer a 65-nm 4600+ model so far). So you cannot be "healthy and rich" with AMD products, it either healthy or rich here.
Memory modules for our testbeds were kindly provided by
Corsair Memory Russia
Stanislav Garmatiuk (email@example.com)
September 12, 2007
Write a comment below. No registration needed!
Copyright © Byrds Research & Publishing, Ltd., 1997–2011. All rights reserved.