AMD Athlon 64 X2 4800+ Anew: AMD Masters 65nm Technology

Without much noise AMD has begun to ship processors based on the 65nm Brisbane core. Does this new core make the company any more competitive against Intel’s Core 2 Duo? Let’s find out now from our new article!

by Ilya Gavrichenkov
01/03/2007 | 01:42 PM

It’s exactly one year since the first processor manufactured with 65nm technology was released. Intel pioneered the use of the new tech process on the borderline of the years of 2005 and 2006 with its Presler core that implemented the NetBurst architecture (for details see our article called First Look at Presler: Intel Pentium Extreme Edition 955 CPU Review). AMD had been traditionally lagging behind its opponent in mastering new manufacturing technologies and was still using its time-tested 90nm tech process with SOI. It could be noted that AMD had virtually exhausted the resources of that technology: it’s now about half a year since we last saw a considerable growth of frequencies of mass-produced CPUs with the K8 micro-architecture.

 

The announcements of Athlon 64 FX processors with frequencies up to 3GHz doesn’t contradict that fact because they are being shipped in very limited quantities and AMD can satisfy the demand by just culling chips capable of working at that clock rate. Thus, the Athlon 64 X2 desktop processor family could only progress further by utilizing new tech processes and AMD mastered mass production of 65nm semiconductor CPU dies, reporting success in early December. Right now Athlon 64 X2 processors on the Brisbane core, manufactured on the new tech process, are emerging in shops.

Besides the new manufacturing technology, AMD has also introduced 300mm wafers. Coupled with the smaller CPU die size, this should reduce the cost of the end-product. Perhaps this will give AMD some room for price maneuvering in its competitive struggle with Intel until faster CPUs with the improved K8L micro-architecture are ready to hit the market. Besides the cost factor, the Brisbane core will obviously allow to produce more energy-efficient processors. Hopefully, the Athlon 64 X2 will now be able to challenge the Core 2 Duo at least in terms of performance per watt. After all, the Brisbane doesn’t bring about any improvements in the micro-architecture and there seems to be no reason why we should expect a performance growth from the new processors.

We don’t want to theorize idly, but are going to test a new CPU from AMD in our labs. This review is about the Athlon 64 X2 4800+ processor based on the Brisbane core.

Closer Look at Brisbane

The heading of this section may be misguiding because we don’t have anything special to say about the new Brisbane core. With this core AMD solved the problem of transitioning to a more progressive tech process, so it is in fact a redesigned version of the Windsor, which is the core of all dual-core Socket AM2 processors manufactured on 90nm technology. It’s in the die size, thermal and electric characteristics that we should look for any differences. This is illustrated by the following table:

The Brisbane core incorporates as many transistors as the Windsor core with the same amount of L2 cache memory does. This indicates the lack of any changes on the micro-architecture level on the transition to the thinner tech process. However, the new technology has helped reduce the area of the Athlon 64 X2 die by over 30%.

Unlike its predecessor, the Brisbane exists in one version only, namely with 512KB of L2 cache per core. The version with a larger amount of L2 cache hasn’t yet been redesigned. It’s clear that the reduction of the manufacturing cost is one of the main problems AMD solves with its 65nm core.

This is also indicated by the frequencies of the CPU models based on the new core. AMD has introduced it into midrange rather than into top-end products first.

Well, it’s not exactly correct to say that the Brisbane doesn’t bring anything new at all. There are indeed certain changes, however negligible they may seem. Particularly, AMD’s engineers have implemented support for fractional frequency multipliers that now change with a step of 0.5x instead of 1x as before. As a result, Brisbane-core CPUs may vary by only 100MHz in their default clock rates. This innovation gives some flexibility to the Athlon 64 X2 series due to the enlarged product assortment.

AMD’s New Series of Dual-Core CPUs for Socket AM2

So, what CPU models currently employ the 65nm core? You’ll get the answer in the following table which shows the current status of the Athlon 64 X2 for Socket AM2 series.

As you can see, the 5000+ model is right now the only one in the Athlon 64 X2 series that can be based on different cores. Yet it’s simple to differentiate the cores in this case – by the typical heat dissipation. CPUs with the new 65nm core have a typical heat dissipation of 65W, which equals that of the Energy Efficient models on the Windsor core (AMD hasn’t been able to offer an energy-efficient version of the Athlon 64 X2 5000+ until now, though).

Thus, the Brisbane core allows AMD to transform its Energy Efficient processors into truly mass products. The 65nm tech process helps reduce the heat dissipation and the TDP of the new mainstream Athlon 64 X2 processors is indeed the same as that of the competing Intel Core 2 Duo series.

The clock rates of the other processors on the Brisbane core do not cross the frequencies of the Windsor-core CPUs, except for the 5000+ model. This is achieved by means of the fractional multipliers available in the new CPUs. The Brisbane-core CPUs have acquired those ratings that used to belong to midrange models with a total of 2MB of L2 cache memory. Such models have been excluded from the series by now, so only the Athlon 64 X2 5200+ and 5600+ models and the Athlon 64 FX series currently have that much of cache.

AMD plans to enlarge the range of dual-core Brisbane-based processors (supposedly in the second quarter of 2007) by releasing models with ratings of 5200+ and 5400+ and clock rates of 2.7GHz and 2.8GHz, respectively. Besides that, the bottom part of the 65nm CPU series is expected to be duplicated with more economical solutions with a typical heat dissipation of 35W.

Slower than Windsor? Brisbane in Synthetic Benchmarks

Same-frequency CPUs on the Brisbane and Windsor cores shouldn’t differ much in performance as they incorporate the same number of transistors and, as AMD claims, don’t differ at all in their micro-architecture. However, there are publications on the Web that argue that point. Let’s check it out.

First, we compared the speed of the computing units of processors on the Brisbane and Windsor cores (the latter had a total of 1MB of L2 cache). For the comparison to be correct, we set the clock rate of both CPUs at 2.4GHz. We tested them in the CPU benchmarks from the SiSoftware Sandra XI suite.

Brisbane 2.4GHz

Windsor 2.4GHz

Sandra XI, Arithmetic ALU

17489

17480

Sandra XI, Arithmetic SSE3

14786

14788

Sandra XI, Multi-Media Integer MMX/SSE

44863

44897

Sandra XI, Multi-Media Floating-Point SSE2

49339

49335

The main computing units of the two CPUs indeed provide the same performance. The difference fits within the measurement error range.

The CPU benchmarks from SiSoftware Sandra XI do not depend on the memory subsystem speed. They are indicative of the “pure” performance of a CPU. But in real-life applications the speed at which the CPU is receiving data from memory has an effect on performance, too. So, we measured the bandwidth and latency of system memory as well as the latency of the L2 cache. We installed dual-channel DDR2-800 memory with timings of 4-4-4-12-1T for this test.

Brisbane 2.4GHz

Windsor 2.4GHz

Sandra XI, L2 Cache Latency, clk

22.7

17.7

Sandra XI, Memory Bandwidth, MB/s

8351

8675

Sandra XI, Memory Latency, ns

107

92

Here’s a strange surprise to you. The L2 cache of CPUs on the new 65nm Brisbane core has higher latency and this increases the latency of the memory subsystem in general. And this ultimately reduces the overall memory subsystem bandwidth.

The discouraging results produced by SiSoftware Sandra XI are confirmed by other synthetic benchmarks of the memory subsystem.

Brisbane 2.4GHz

Windsor 2.4GHz

CPU-Z, L2 Cache Latency, clk

20

12

CPU-Z, Memoy Latency, clk

115

108

ScienceMark 2.0, L2 Cache Latency, clk

20

13

ScienceMark 2.0, Memory Latency, clk

114

106

ScienceMark 2.0, Memory Bandwidth, MB/s

7619

8202

EVEREST 2006, Memory Read, MB/s

7816

8044

EVEREST 2006, Memory Write, MB/s

6833

6932

EVEREST 2006, Memory Copy, MB/s

7914

8152

EVEREST 2006, Memory Latency, ns

51

48.7

So, there can’t be any doubt: the L2 cache has become slower in the new core for Athlon 64 X2 CPUs. As a result, the new CPUs work slower with data in memory than their 90nm predecessors, which may show up in real-life applications, too. But the organization of the L2 cache hasn’t changed: 16-way associativity with a line length of 64 bytes.

So, we should seek elsewhere for the root of the problem. AMD has commented that the latency of the L2 cache has increased as a consequence of the engineers having left a reserve for enlarging the cache in the future. This doesn’t sound convincing to us, however. First, AMD’s plans don’t contain any information about enlarging the L2 cache even on the transition to the K8L micro-architecture. Second, Windsor-core CPUs with a 2x1MB L2 cache do not differ in cache performance from their counterparts that are equipped with a 2x512KB L2 cache. So, it is not yet clear to us why the speed parameters of the cache have changed.

The increased latency of the cache memory is not the only problem that can have a negative effect on performance of Brisbane-core CPUs. Another problem is about the fractional CPU frequency multipliers – the real memory frequency has been reduced in some modes because the default CPU frequencies now change with a step of 100MHz. In CPUs with the K8 micro-architecture the memory frequency is actually based on the CPU frequency and an integer divider. We’ve met this problem before, but it has grown worse with the new CPUs. To illustrate our point, here is a table that shows the real memory frequency in the different modes of the memory controller integrated into the CPU.

Processor frequency, MHz

2000

2100

2200

2300

2400

2500

2600

2800

DDR2-800

800

700

733

767

800

714

743

800

DDR2-667

667

600

629

657

600

625

650

622

DDR2-533

500

525

489

511

533

500

520

509

That’s not a catastrophe, of course. There are very poor memory modes with CPUs that have integer frequency multipliers, too. Yet you should be aware of this thing because the real memory frequency is often much lower than the expected one with CPUs that have fractional multipliers. This has a negative effect on the overall system performance, of course.

Athlon 64 X2 4800+

Now we can have a closer look at the Athlon 64 X2 4800+ processor on the 65nm Brisbane core that we’ve got for our tests.

As the photograph shows, the Athlon 64 X2 4800+ has an ordinary appearance. The ADO4800IAA5DD marking suggests that AMD formally puts this model into the Energy Efficient class which is expectable considering its typical heat dissipation of 65W. It has diminished due to the thinner tech process as well as to the lower core voltage (it is 1.25-1.35V in general, and 1.35V with our particular CPU).

The informational program CPU-Z correctly identifies the new core and its basic characteristics.

Take note of the 2.5GHz clock rate and the fractional multiplier of 12.5x. This looks somewhat unusual.

The screenshot doesn’t tell you the core stepping, but it is actually G1. You can check this out at www.amdcompare.com.

Testbed and Methods

We benchmarked our Athlon 64 X2 4800+ on the 65nm Brisbane core in comparison with an Athlon 64 X2 5000+ and an Athlon 64 X2 4600+ that were based on the older 90nm Windsor core. Besides them, we took an Intel Core 2 Duo E6400 processor that is somewhat cheaper than the Athlon 64 X2 4800+.

We used the following hardware for our tests:

We selected the highest-performance settings in the mainboards’ BIOS Setup.

Performance

SYSMark 2004 SE: Overall Performance

Our first benchmark shows that the flaws in the new 65nm CPU we’ve discussed above do not have a big effect on its performance. Yet you may note that the speed of the Athlon 64 X2 4800+ is closer to that of the Athlon 64 X2 4600+ than of the Athlon 64 X2 5000+. Representing the opposite camp, the Core 2 Duo E6400 is considerably faster than any of the tested Athlon 64 X2 in SYSMark 2004 SE, although is cheaper even than the Athlon 64 X2 4600+.

Futuremark: Synthetic Benchmarks

There are no surprises here. Being slower with memory due to its lower frequency and to the higher latency of the L2 cache, the Athlon 64 X2 4800+ anyway fits exactly within the results of the Athlon 64 X2 5000+ and Athlon 64 X2 4600+.

Gaming Applications

It is in gaming applications that the defects in the memory subsystem of the Brisbane-core CPU show up. Contradicting their respective ratings, the new Athlon 64 X2 4800+ is slower than the older Athlon 64 X2 4600+ by about 3-7%.

Besides the two games, we ran two benchmarks based on the Valve Source engine that is expected to be used in the upcoming Half-Life 3 . One benchmark shows the speed of processing the environment physics and the other measures the speed of creating lighting maps.

Notwithstanding their native support for multi-core processors, game engines of the future don’t seem to differ much in their CPU preferences from today’s engines.

Video and Audio Encoding

These diagrams don’t have anything new to tell us. The Athlon 64 X2 4800+ CPU on the new Brisbane core is a little faster than the Athlon 64 X2 4600+ on average. Note also that the new CPU cannot challenge the Core 2 Duo E6400, although costs more (by about $50, according to the official price-list).

Image and Video Editing

These tests produce the same picture of performance as the previous ones.

Professional OpenGL and Final Rendering

The results produced by the popular professional suite 3ds max 9 aren’t surprising, either. It’s clear that the memory subsystem speed doesn’t affect the overall performance much at final rendering. This influence is higher when you’re working in the projection windows and becomes a crucial factor when you’re working with shader code.

Overclocking

Brisbane-core processors haven’t surprised us with their performance in most applications, even though their L2 cache is slower than that of the Windsor-core models. But besides the default performance, PC enthusiasts may be interested in an overclockability check because a transition to a thinner tech process usually pushes the CPU frequency bar higher. This rule may not apply to the Brisbane, though, as the default clock rate of the new CPUs, declared by AMD, is lower than the frequency of the CPUs on the older 90nm core.

We tried to check this out by overclocking our CPU. We cooled it with an air cooler Zalman CNPS9500 LED. The frequency multiplier of the Brisbane-core Athlon 64 X2 4800+ being limited by the default 12.5x from above, this CPU has to be overclocked in the usual way, by increasing the clock generator frequency. To avoid problems with the other subsystems of the testbed, we reduced the multipliers for the memory and HyperTransport buses (the latter connects the CPU with the mainboard’s SPP).

First, we tried to overclock the CPU at its default voltage of 1.35V. We managed to increase the clock-gen frequency by 15%, to 230MHz.

The system with the Brisbane-core CPU overclocked to 2.87GHz was perfectly stable. This was a good result, but no record in comparison with what you could achieve with Windsor-core CPUs. Then we increased the voltage of our Athlon 64 X2 4800+ to 1.6V. This 18% increase in voltage produced a rather small effect: the maximum stable CPU frequency grew only 4% higher.

Thus, the maximum of frequency we reached without any special cooling methods was 3.0GHz.

In other words, the Brisbane core didn’t provide any advantages over the Windsor in terms of overclockability, either (AMD currently offers 90nm processors for which 3.0GHz is the default frequency). So, you shouldn’t as yet expect any overclocking breakthroughs from AMD’s transition to 65nm tech process.

Power Consumption Tests

After the mediocre results of the Athlon 64 X2 4800+ processor on the 65nm Brisbane core in our performance and overclockability tests, it is now clear that it can only prove to be superior to its predecessors in efficiency because the typical heat dissipation of every Athlon 64 X2 on the new core is 65W whereas the typical heat dissipation of CPUs on the 90nm Windsor core is 89W. However, AMD has been offering Energy Efficient processors on the 90nm Windsor core since the last half of the summer. Such CPUs have a typical heat dissipation of 65W and it’s not quite clear without additional tests how they compare with the new 65nm CPUs. So, we added an Energy Efficient Athlon 64 X2 4600+ on the 90nm Windsor core into our power consumption tests.

In these tests we use the same platform but with different CPUs. We measure the current passing through the CPU power circuit – it is indicative of the power consumption of the CPU (but without considering the efficiency of the CPU power converter). The CPU is loaded up by running S&M and Intel Thermal Analysis Tool. We activate the power-saving technologies AMD Cool’n’Quiet and Enhanced Intel SpeedStep for the test. Here are the results:

It’s interesting: the Intel Core 2 Duo E6400 is more economical than AMD’s solutions under full load but consumes more power when idle. Comparing AMD’s CPUs, the 65nm model seems to provide a better performance-per-watt ratio than the ordinary Athlon 64 X2, but the Energy Efficient Athlon 64 X2 4600+ consumes less than the new CPU under full load.

So, AMD’s new CPUs aren’t so much better than the older ones in their energy-related characteristics. AMD has already managed to achieve the same power consumption with the older Windsor core. We must acknowledge, though, that they had to cull dies capable of working at reduced voltage then while the new Brisbane core seems to achieve the same purpose without any special tricks. Moreover, the maximum frequency of AMD’s Energy Efficient CPUs on the Windsor core is 2.4GHz whereas the new core can work at 2.6GHz (and is expected to work at 2.8GHz in the future), having a typical heat dissipation of 65W, too.

Conclusion

The new 65nm Brisbane core can’t provoke a revolution on the CPU market. From an ordinary user’s point of view, the processors based on the new core differ but little from their predecessors. The new core doesn’t feature any improvements on the micro-architecture level and its frequency potential doesn’t differ much from that of the older Windsor core.

So, it is AMD itself that will profit most from the Brisbane. With the smaller CPU die area and larger semiconductor wafers, the cost of dual-core CPUs is now reduced whereas the introduction of fractional frequency multipliers allows to shape up a more flexible CPU series. Having mastered the 65nm tech process prior to launching its CPUs with the improved K8L micro-architecture, AMD has got a chance to make the manufacturing process polished off and ready for the moment the company’s fate will be decided.

As for consumer properties of the new CPUs, the Brisbane core does offer something positive. First of all, the 65nm CPUs have become more economical (if you don’t compare them with the older models from the Energy Efficient series) – all the new CPU models on the Brisbane core will fit within a TDP of 65W. Second, the new CPUs may be somewhat better to deal with for an overclocker.

But is this enough to make the new CPUs appealing in the eyes of the rapidly dispersing crowd of AMD fans? We guess, not. Especially as the Brisbane’s drawback – it works slower than the previous core with data in memory– may outweigh the good points mentioned.