AMD was faced with a tradeoff during the development of the dual core Athlon 64 X2. In order to maintain backwards compatibility with earlier Socket-939 motherboards, they could not change the pinout of their dual core processors. While maintaining the same pinout resulted in the ability to upgrade virtually any Socket-939 platform to a dual core Athlon 64 X2, it meant that the dual core processors were left with no more memory bandwidth than their single core counterparts. The single-core Socket-939 Athlon 64s feature a 128-bit wide DDR memory controller, which when operating at DDR400 speeds, it gives the A64 a maximum of 6.4GB/s of memory bandwidth. Sharing the same memory controller, the dual core Athlon 64 X2s also feature the same 6.4GB/s of memory bandwidth, despite the fact that there are now twice as many cores vying for the same amount of memory bandwidth.

Luckily for AMD, the single core Athlon 64 was not very memory bandwidth limited, and thus, the move to dual core still allowed AMD to scale relatively well. In fact, based on the results that we saw in our Athlon 64 X2 3800+ review, AMD continues to consistently scale better from one to two cores than Intel, despite the reduction in memory bandwidth per core.

Meanwhile, AMD quietly introduced a handful of new memory dividers in the latest revisions of their Athlon 64 and Athlon 64 X2 processors. These new memory dividers allow for memory clock speeds above DDR400 to be enabled without overclocking the Hyper Transport bus. The beauty of these new memory dividers is that owners of faster-than-DDR400 memory can take advantage of the extra bandwidth offered by their modules, without overclocking their CPUs or the rest of their system.

Last month, we took a look at the performance benefit, or honestly, the lack thereof with using higher bandwidth memory and Athlon 64/X2 processors. For the most part, we saw a 0 - 3% improvement in real world performance, with the vast majority of benchmarks showing us a 0 or 1% increase in performance, thanks to the higher bandwidth memory. There were some isolated cases where having more memory bandwidth translated into higher performance, in particular things like video encoding, gaming and heavy multitasking environments, but for the most part, the performance gains were negligible.

The performance gains in video encoding and gaming were to be expected, and we theorized that there would be some significant gains in multitasking environments. In a multitasking environment, particularly with an Athlon 64 X2, the overall memory bandwidth requirements of the two combined cores should be at their peak, well above and beyond the demands of a single-core Athlon 64. We saw this in our original article where one of our heavier multitasking tests yielded a 6.5% increase in performance when using DDR480 with an Athlon 64 X2 4800+. At the same time, some of our lighter multitasking tests yielded absolutely no performance increase when paired with higher bandwidth DDR memory. So, the point of this article is to find out if multitasking Athlon 64 X2 owners can benefit any more than single-core users from employing these new faster-than-DDR400 memory speeds.

Given the very specific nature of this article, we’re only going to be focusing on one processor - the Athlon 64 X2 4800+. As we found in our last piece, slower X2s weren’t impacted any differently than the fastest of the bunch, so anything we find here should be just as applicable in the real world to all other X2 processors.

We also only focused on two memory speeds: the base DDR400 and the fastest possible setting on the 4800+, DDR480. The details of how to select these speeds and the hardware we used to do so can be found in our first article .

Multitasking Office Performance

Our first test is actually the scripted Multitasking Winstone 2004 test. We’ve used this test in the past, and it serves as an excellent example of relatively light general use multitasking performance. The test consists of three parts, all of which are described below:

"This test uses the same applications as the Business Winstone test, but runs some of them in the background. The test has three segments: in the first, files copy in the background while the script runs Microsoft Outlook and Internet Explorer in the foreground. The script waits for both foreground and background tasks to complete before starting the second segment. In that segment, Excel and Word operations run in the foreground while WinZip archives in the background. The script waits for both foreground and background tasks to complete before starting the third segment. In that segment, Norton AntiVirus runs a virus check in the background while Microsoft Excel, Microsoft Project, Microsoft Access, Microsoft PowerPoint, Microsoft FrontPage, and WinZip operations run in the foreground."

We’ve been playing around with multitasking performance tests for several months now, and have found that even scripted tests like the Multitasking Winstone test require a lot of work to get to produce repeatable results. The problem mostly boils down to making sure that all of the tasks executing simultaneously do so in the exact same manner, every single time, across all platforms, CPUs and other configuration changes. Honestly, doing so is very difficult and it often requires far more benchmarking runs than we are used to performing for most of our other tests. But at the end of the day, it is possible to get results that do make some sort of sense, and spending a great deal of time with Multitasking Winstone and our own home-brew tests, we have done just that.

Multitasking Winstone	DDR400	DDR480	% Improvement
Test 1	2.21	2.37	7.2%
Test 2	2.94	3.05	3.7%
Test 3	4.82	4.88	1.2%

The first test proved to be the most impressive out of the bunch, showing a 7.2% increase in performance over stock DDR400. Note that a 7.2% performance advantage is greater than what we’d see when going from an Athlon 64 X2 4400+ to a 4800+.

The second test still produced reasonably good results, showing a 3.7% increase in performance. The third and final test shows that not all situations will yield a tangible performance increase.

Although it is a canned benchmark, Multitasking Winstone 2004 gives us a very good idea of what is to come. But in order to truly find out if higher bandwidth memory is worth it for Athlon 64 X2 owners, we turned to some of our own home-brew multitasking benchmarks.

Multitasking with Adobe Photoshop CS

Intel has done a great job of putting together a number of benchmarks that put real applications in real world usage scenarios. Despite the validity of most of their benchmarks, we don’t use them in our CPU comparisons because of the fact that they are scripted by Intel. But, since we’re not doing an AMD vs. Intel comparison here today, we made use of one of them in particular - their Adobe Photoshop CS benchmark.

Usually, Photoshop benchmarks consist of performing every single filter on an image and recording the time taken by each filter. Intel’s benchmark performs a bunch of filters and image manipulations much like a user would, and keeps track of the total time that the script took to complete.

First off, let’s take a look at the performance of DDR400 vs. DDR480 without anything running in the background, just Photoshop CS:

Time in Seconds (Lower is Better)	DDR400	DDR480
Adobe Photoshop CS	63.263s	63.028s

Although DDR480 is slightly faster than DDR400, it isn’t a tangible performance advantage (less than 1%).
But now to spice things up a bit, let’s look at Photoshop CS performance when we’re listening to a MP3 and encoding a video at the same time. For this particular test, we start up Windows Media Encoder 9 and iTunes, begin encoding our test video, then start playing our single MP3 on repeat and start the Photoshop CS script 5 seconds later.

Time in Seconds (Lower is Better)	DDR400	DDR480
Adobe Photoshop CS + MP3 + WME9	105.773s	98.851s

The end result is far more tangible this time around, with DDR480 completing the Photoshop task in 6.5% less time than the same system with DDR400.

We used our Windows Media Encoder test here because in our original article, the WME test itself saw no benefit from DDR480, much like Photoshop CS. But putting the two together and tossing in some background music changed the picture; it changed things enough that we’d actually recommend using higher bandwidth memory in this case.

Multitasking with Doom 3

Gaming performance also had a reasonable improvement in our first article, with DDR480 realizing a 4% increase in performance over standard DDR400. We paired the Doom 3 benchmark with a WME9 encoding task and watched the performance go up even further:

Doom 3 Performance (fps - higher is better)	DDR400	DDR480	% Improvement
Doom 3 w/ background WME9	85.1	93.3	9.6%

Pairing up two memory-bandwidth intensive tasks, we see a very noticeable 9.6% increase in performance. It’s generally said that you really start to notice performance improvements of about 10%, and with this particular situation, we are right there.

Multitasking with 3dsmax

Not all multitasking scenarios proved to be so happy with the added memory bandwidth, and this next test is an example of just that.

For this test, we combined a 3dsmax 6 render task with MP3 playback as well as our standard Firefox browser test (with 12 open tabs, featuring a decent amount of flash).

3dsmax Render Time in Seconds (lower is better)	DDR400	DDR480
3dsmax + MP3 + Firefox	99.828s	98.813s

The performance improvement that DDR480 offered was basically next to nothing, which shows that not all multitasking environments are particularly memory bandwidth intensive.

Multitasking with File Compression

Compressing and decompressing files is something we all do, and watching H.264 encoded movies is the latest and greatest way to stress your system. So, what about doing both at the same time?

For this test, we played the Batman Begins H.264 trailer and ran the built-in WinRAR benchmark. We measured the average compression rate in MB/s:

WinRAR Compression Rate in MB/s (Higher is Better)	DDR400	DDR480	% Improvement
WinRAR + H.264 Decode	24.5MB/s	25.7MB/s	4.6%

These two tasks combined yielded a 4.6% increase in performance for DDR480 over DDR400.

Video Encoding + Decoding

Who says that video must be encoded and decoded separately? For this next test, we played the Batman Begins H.264 trailer while performing our standard DivX 6 encode test in the background.

All results were measured in average DivX 6 frames encoded per second:

DivX Encode Rate in Frames per Second (Higher is Better)	DDR400	DDR480	% Improvement
DivX Encode + H.264 Decode	35.7 fps	38.4 fps	7.6%

Combining two very memory bandwidth intensive tasks nets a 7.6% increase in performance; once again, not bad at all.

PAR2 + Encoding

Creating and using parity files with archives is another very CPU and memory intensive task. We fired up Quickpar and restored a single file from an archive and its set of PAR2 files; alongside the PAR2 process, we performed a couple of tasks.

First, let’s look at the archive reconstruction throughput without any other tasks:

PAR2 Archive Reconstruction Rate in MB/s (Higher is Better)	DDR400	DDR480	% Improvement
PAR2	931MB/s	941MB/s	1%

DDR480 only offers a 1% performance advantage here, but now let’s add MP3 playback in the background:

PAR2 Archive Reconstruction Rate in MB/s (Higher is Better)	DDR400	DDR480	% Improvement
PAR2 + MP3 Decode	904MB/s	918MB/s	1.5%

The DDR480 performance advantage jumps to 1.5%, but now, let’s add a H.264 encode on top of that:

PAR2 Archive Reconstruction Rate in MB/s (Higher is Better)	DDR400	DDR480	% Improvement
PAR2 + MP3 + H.264 Encode	843MB/s	880MB/s	4.4%

And now, we have a 4.4% performance advantage.

You can see how the performance differences scale according to the tasks that you’re pairing together.

Watching Movies while you Work

For our final tests, we combined a viewing of the Batman Begins H.264 trailer with the standard Business and Multimedia Content Creation Winstone 2004 tests. The idea behind these tests was to take a look at some of the lighter multitasking loads, combining video playback with a variety of applications.

Winstone 2004 Performance (Higher is Better)	DDR400	DDR480	% Improvement
Business Winstone 2004 + H.264 Playback	19.9	20.4	2.6%
Multimedia Content Creation Winstone 2004 + H.264 Playback	39.6	39.9	‹ 1%

As expected, the performance boosts from lighter multitasking scenarios are next to nothing. Surprisingly enough, Business Winstone saw more of a performance boost at 2.6%, while MMCC Winstone showed us a less than 1% increase in performance.

Final Words

The outlook for using faster-than-DDR400 memory is far less bleak when you consider multitasking scenarios and the Athlon 64 X2. While not all of the scenarios that we’ve put forth showed significant performance improvements, there were more areas where we did see performance gains than not this time around.

So, while our recommendation for not needing higher bandwidth DDR memory remains as is for single core Athlon 64 owners, if you happen to be a heavy multitasker, you’ll find that the Athlon 64 X2 can benefit from higher speed memory in scenarios like the ones that we’ve outlined here today.