Posted:2007-01-09 By cpu review Number of View:129430
AMD ATHLON 64 X2 4800+ BRISBANE
REVIEW BENCHMARK AND OVERCLOCKING
By :cpu review
Posted:2007-01-09
xtreview is your : Video card - cpu - memory - Hard drive - power supply unit source
From the first processors released with the technological process 65 nm passed exactly one year. Intel company represented on boundary of 2005 and 2006 its core presler, based on the micro-architecture netBurst. But AMD, traditionally lagging behind its basic competitor on the speed of introduction with new production technologies, entire 2006 lived on the customary and well fixed by 90 nm technical process with application SOI. And, as it was sufficiently long time ago noted the possibility of this AMD technology it already limited : There is no considerable increase in the clock frequencies of CPU with the micro-architecture K8 for more than half a year.
It is obvious that further progress of the desktop processor athlon 64 X2 will be possible only with the introduction of new technological processes. And, fortunately AMD finnaly approached the mass production of 65 nm semiconductor processor crystals. At present in some stores already begin to appear processors athlon 64 X2, at basis of which lie the new cores brisbane, released with the application of a new technological process.
we should note that together with the beginning of the new technology application ,AMD began the introduction of 300 mm base layers, which together with the reduction in area the core of processor, must lead to a reduction in the prime cost of finished production. Probably, this fact will give boost for the new price vs the competitive activity from Intel. Besides reduction in the prime cost the core brisbane, obviously will make possible for AMD to release more economical from the point energy consumption and heat emission CPU. It is desirable to hope that now Athlon 64 X2 can compete with Core 2 Duo at least from the side performance per watt . Indeed there are no micro-architectural improvements into Brisbane, and it means we should not wait forincrease in the performance from new processors .
Brisbane in details
Despite the name of this division, some special details about the new core brisbane to report is sufficiently complicated. The creation of this the core by AMD posed the only problem of the mastery of more advanced technological process. SO, Brisbane is only commonplace of well familiar to us core windsor, which lies today at the basis of all dual cores socket AM2 processors, produced on 90 nm technology. Differences should be searched for only in the area core, the thermal and electrical characteristics.
Brisbane
Windsor
Windsor
Ratings of processors
4000+ - 5000+
3800+ - 5400+
5200+ - 5600+
clock frequency
2.1 – 2.6 GHz
2.0 – 2.8 GHz
2.6 – 2.8 GHz
Micro-architecture
K8
K8
K8
Technological process
65 nm
90 nm
90 nm
Processor Socket
Socket AM2
Socket AM2
Socket AM2
Processor system bus
8 Gigabite/s hyperTransport
8 Gigabite/s hyperTransport
8 Gigabite/s hyperTransport
Supported types of memory
DDR 2-800/667/533 SDRAM
DDR 2-800/667/533 SDRAM
DDR 2-800/667/533 SDRAM
Number of transistors
153.8 mln.
153.8 mln.
227.4 mln.
Crystal area
126 sq. mm
183 sq. mm
230 sq. mm
L1 cache instructions
2 x 64 kbyte
2 x 64 kbyte
2 x 64 kbyte
L1 cache data
2 x 64 kbyte
2 x 64 kbyte
2 x 64 kbyte
L2 cache
2 x 512 kbyte
2 x 512 kbyte
2 x 1024 kbyte
Technologies of energy-economy
CoolandQuiet
CoolandQuiet
CoolandQuiet
the 64- bit expansions
AMD64
AMD64
AMD64
Technology of virtualization
Pacifica
Pacifica
Pacifica
SIMD instruction
SSE3
SSE3
SSE3
The number of transistors in the core brisbane is equal to the number of transistors in core windsor, which possesses analogous volume L2 cache. This, strictly indicates the absence of changes at the micro-architectural level upon transfer to the more advanced technical process. In the same time the application of newer technology production made possible to decrease the area core of processor athlon 64 X2 more than by 30%.
It is important to note that Brisbane in contrast to predecessor exists only in one version with the reduced to 512 kbyte on core cache in the second level. Version with large volume do not excit at the moment .
The same indirectly indicate the frequencies of processors, based on the new core. AMD began its introduction by the upper models.
However, we can not speak that the appearance Of brisbane has not introduced any update. Namely, the engineers OF AMD realized in it the support of the fractional coefficients which change now not with discretion 1x, as earlier, but through 0.5x. as a result, the processors, which are based on Brisbane, can have regular clock frequencies, which differ from other only by 100 MHz.
New dual core socket AM2 processors AMD
In what processors it is possible to meet the today new core, produced on the technological process with the standards production 65 nm? Answer to this question is in the following table, in which is reflected the current state of processors athlon 64 X2 socket AM2.
Processor
The clock frequency
Core
L2 cache
TDP
Athlon 64 X2 5600+
2.8 GHz
Windsor
2 x 1 Mbytes
89 W
Athlon 64 X2 5400+
2.8 GHz
Windsor
2 x 512 kbyte
89 W
Athlon 64 X2 5200+
2.6 GHz
Windsor
2 x 1 Mbytes
89 W
Athlon 64 X2 5000+
2.6 GHz
Windsor
2 x 512 kbyte
89 W
Athlon 64 X2 5000+
2.6 GHz
Brisbane
2 x 512 kbyte
65 W
Athlon 64 X2 4800+
2.5 GHz
Brisbane
2 x 512 kbyte
65 W
Athlon 64 X2 4600+
2.4 GHz
Windsor
2 x 512 kbyte
89/65 W
Athlon 64 X2 4400+
2.3 GHz
Brisbane
2 x 512 kbyte
65 W
Athlon 64 X2 4200+
2.2 GHz
Windsor
2 x 512 kbyte
89/65 W
Athlon 64 X2 4000+
2.1 GHz
Brisbane
2 x 512 kbyte
65 W
Athlon 64 X2 3800+
2.0 GHz
Windsor
2 x 512 kbyte
89/65/35 W
The contemporary state of athlon 64 X2 is such, that different cores can be used only in one model processor, with rating 5000+. But it is obvious that to distinguish core in this case will be sufficient simply – on the typical heat emission. level in processors with the new 65 nm core is equal to 65 W and coincides with the analogous characteristic Of energy efficient models with core windsor. However, until now, AMD could not propose economical solution for athlon 64 X2 5000+.
Thus, appearance of core brisbane makes possible FOR AMD to convert its Of energy efficient processors into the properly mass solution: 65 nm technical process give the possibility to this producer to decrease the heat emission of its CPU . Actually, TDP of the new mass processors athlon 64 X2 was equaled with the typical heat emission of competing CPU intel core 2 Duo.
Subsequently AMD company AMD plans to enlarge the assortment of dual core processors with 65 nm core brisbane due to the production of the corresponding models with ratings 5200+ and 5400+, whose clock frequencies will be 2.7 and 2.8 GHz. Furthermore, it is expected that over the long term the lower part of CPU with 65 nm core will be doubled by more economical proposals, whose typical heat emission will be lowered to 35 W.
Brisbane is slower than Windsor? Synthetic tests
In the first turn we made the decision to compare the speeds of the computational blocks for processors based on the new and old core. For this we compared the performance of two CPU: based on core brisbane and being been based on core windsor (and possessing L2 cache with the total volume of 1 Mbytes). So that this comparison would make sense, clock frequency of both processors was advanced in same value of – 2.4 GHz. In testing we adapted the processor benchmarks, entering the test packet Of siSoftware sandra XI.
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 operating speed of the basic computational units of processor cores as was expected, proved to be identical.
Processor tests from SiSoftware sandra XI are good fact that their results do not depend on the speed of memory subsystem operation , they characterize only the clean performance of processor. in real applications effect on the operating speed has naturally the impact of memory . Therefore, separately we have measured the memory speed, and also its latency and latency L2 cache. In this test by us was used two-channel DDR 2-800 memory, which works with latency 4- 4- 4- 12-1T.
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
But here with the measurement memory speed we got unexpected result. As show the obtained results, L2 processor cache with the new 65 nm core brisbane has higher latency, that result in increase in the latency of the entire memory subsystem . But in the final sum this gives also reduction in the integral memory subsystem speed .
The obtained into SiSoftware sandra XI results are confirmed by other synthetic tests utility .
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
we cannot Doubts more: L2 cache in the new core of processors athlon 64 X2, produced on the new 65 nm technological process, was slowed down. In summation, new processors work with the data more slowly than their 90 nm predecessors, which, naturally, will be reflected also in performance in real applications.
It should be noted that organization of cache in the second level in this case did not change, it has 16 regions associative property and lengths line of 64 bytes.
The commentary of AMD representatives in regard to this says, that the latency L2 cache increased as a result of reserve for increase in the volume of cache made by engineers of company. But the in our opinion, this explanation appears so convincingly. First, the current plans AMD do not contain information about the planned increase in the volume cache in the second level, even upon transfer to the micro-architecture K8L. Secondly, the processors, built on the core windsor with L2 cache total volume of 2 Mbytes, demonstrate no differences in the speed of cache in comparison with analogs, with the reduced to 1 Mbytes cache. In other words, the question about the reasons for change in the speed parameters of cache, in our opinion still open.
There is another unpleasant moment, connected with the use for the formation of the frequency CPU from fractional coefficients. The alternation frequencies with step of 100 MHz involved reduction in the real frequency of memory in some regimes. This is caused by the fact that for its formation in the processors with the micro-architecture K8 is used the frequency of processor and integral dividers. A similar problem floated earlier....
In order to illustrate this fact, let us give the table, in which are indicated the real frequencies of memory with the selection of different regimes for the functioning built-in in CPU memory controller.
Processor Frequency in MHz
2000
2100
2200
2300
2400
2500
2600
2800
DDR 2-800
800
700
733
767
800
714
743
800
DDR 2-667
667
600
629
657
600
625
650
622
DDR 2-533
500
525
489
511
533
500
520
509
Situation, certainly is not catastrophic, among processors with the entire coefficients also are encountered very unsuccessful versions of memory. But we should not forget this aspect , since for CPU with the fractional coefficients the real frequency of memory sufficiently frequently differed from the expected smaller side, which negatively affects the general performance of system.
Test processor: Athlon 64 X2 4800+
Approached time in detail to introduce the processor athlon 64 X2 4800+, the based on 65 nm brisbane, which proved to be in our laboratory :-).
As it is evident on the photograph, the exterior view in Athlon 64 X2 4800+ is completely standard. Marking ADO4800IAA5DD makes possible to conclude that formally AMD made this processor in the class Of energy efficient, that it is not amazing, taking into account its typical heat emission in 65 W. Reduction in the heat emission in this case is achieved by both application of more advanced technological process and by decrease in the supply voltage core, which comprises in general case 1.25-1.35 v, and for our concrete processor it was equal to 1.35 v.
Diagnostic utility CPU- Z adequately determines new core and its fundamental characteristics.
You will turn attention to the frequency of 2.5 GHz and fractional coefficient 12.5x: . Stepping core on the screen shot it is not indicated, but it is called G1, this information can be obtained on the site amdcompare.com.
How we tested
The performace of processor athlon 64 X2 4800+, based on core brisbane, produced on the technological process with the standards production 65 nm, we compared it with speed of processors athlon 64 X2 5000+ and Athlon 64 X2 4600+, which are based on older 90 nm core windsor. Furthermore, we selected competing CPU from Intel core 2 Duo e6400, whose cost is somewhat lower than athlon 64 X2 4800+.
In the composition of test systems was used the following equipment :
ASUS M2N32-SLI deluxe (Socket AM2, NVIDIA nForce 590 SLI).
Memory: 2048MB DDR 2-800 SDRAM (Mushkin XP 2-6400PRO, 2 x 1024 MB, DDR 2-800, 4- 4- 4- 12-1T).
Graphic card: PowerColor x1900 XTX 512MB (PCI- E x16).
Disk subsystem: Maxtor maXLine III 250GB (SATA150).
Operating system: Microsoft Windows XP SP2 wth DirectX 9.0c.
Performance
SYSMark 2004 SE: general performance
The first tests in real applications show that all new processor with 65 nm core, about which was said above in the practice does not influence the performance. Although, it is necessary to note that the speed Of athlon 64 X2 4800+ is more strongly towards to the results Of athlon 64 X2 4600+, than the performance Of athlon 64 X2 5000+. Well let us note that Core 2 Duo e6400, with the cost inferior even than Athlon 64 X2 4600+, into SYSMark of 2004SE perceptibly passes all those participating athlon 64 X2.
Futuremark: Synthetic tests
There are no unexpected contingencies on the graphs. the slower work with memory because of its lower frequency and higher latency of L2 cache, the results Of athlon 64 X2 4800+ fall between the speed Of athlon 64 X2 5000+ and Athlon 64 X2 4600+.
Gaming applications
Problems in the CPU memory subsystem based on core brisbane, show up in gaming applications. here new Athlon 64 X2 4800+ is inferior in speed to the old athlon 64 X2 4600+, moreover the deference is in region 3-7%.
After two real games we included results of two benchmarks based on the engine Valve source, which will be used in the future game half-Life . The first benchmark evaluates the speed of the physics operation environment, and the second the speed of the construction maps illumination.
It is obvious that the generic efficiency of future game engine , the in spite of support to the multicore, will not begin to qualitatively change the picture, observed in contemporary games.
Coding video and audio
Looking at the diagrams, which show the performance of processors during coding of video and audio file, we hardly will be able to make some new conclusions. The performance of processor athlon 64 X2 4800+, based on the new core brisbane, occurs on the average only higher than athlon 64 X2 4600+. this CPU can not compete with Core 2 Duo e6400 , in spite of its higher cost (according to the official Price- list approximately $50).
Editing images and video montage
Analogously appears situation with image processing and nonlinear videomontage.
Professional OpenGL and final rendering
With final rendering, the speed of the memory subsystem do not affects the performance a lot . However, with the work in windows projection this dependence increases.
Overclocking
Performance in the majority of applications was not surprises of processors on core brisbane even with lower performance in comparison with processors core windsor. But besides the performance we was interested in overclocking potential. usually passage to the new more contemporary technological process increases the maximum clock frequencies of processors. However, in the case of Brisbane situation is not so , since the regular frequency of new CPU, , it proved to be less than the frequency of processors which use older 90 nm core.
Cooling processor during these experiments was done by air cooler zalman CNPS9500 LED. Since the coefficient Of athlon 64 X2 4800+ on core brisbane is bounded above by regular value of 12.5x, to overclock such CPU it is necessary to use custom method – increase the frequency of clock generator. In this case in order to avoid problems with the unfitness for work with other subsystems , we reduced coefficient for the memory frequency and for the system bus hyperTransport, connecting processor with SPP motherboard.
In the first turn we took an attempt to overclock processor without increase in the supply voltage, i.e., with 1.35 we succeeded to increase the frequency by 15%, to 230 MHz.
In this state the system based on the overclocked to 2.87 GHz processor with core brisbane, demonstrated complete stability. This is good result;. Following this we decided to try to overclock Athlon 64 X2 4800+, after increasing his voltage to 1.6 v. but The maximum clock frequency, with which the processor preserved its stability, was increased in this case only by 4%.
Thus, the maximum frequency, which we succeeded in obtaining with the overclock without special methods of cooling, was 3.0 GHz.
In other words, the core brisbane demonstrated no special advantages over Windsor, also, from the point overclocking possibilities. For example, at present are accessible processors AMD with 90 nm core, 3 GHz for which it is default frequency. In other words,it does not follow to connect the transfer with AMD company of production CPU on 65 nm technological process in some special overclocking .
Testing the energy consumption
After the ordinary results shown by processor athlon 64 X2 4800+, based on 65 nm core brisbane, in the tests performance and with overclock, it becomes obvious that to exceed its predecessors it will be able only from the point of efficiency. This as the minimum, the value of typical heat emission for all Athlon 64 X2 with the new core is in 65 W, while the typical heat emission of processors with 90 nm core windsor is equal to 89 W. however this assertion needs some reservations. fact that from the second-half of summer AMD proposes Energy efficient processors, based on 90 nm core windsor, whose typical heat emission is reduced to 65 W. But as against the background of old Energy efficient athlon 64 X2 can appear processors with the new core, produced on 65 nm technology, without additional tests it is incomprehensible. In this connection into the number of processors, which were examined by us during testing of energy consumption, into the number of other was added Energy efficient athlon 64 X2 4600+, based on 90 nm core windsor.
During the tests we have done the comparison of the amounts of required power , measured in analogous platforms by different processors. For creating load on processors were used the utilities S&M and Intel thermal analysis tool. The energy-saving technologies AMD coolnQuiet and Enhanced intel speedStep during the tests were activated.
The results of measurements are given on the graphs.
Picture is completely curious. Let us immediately focus attention on the fact that processor core 2 Duo e6400 demonstrates the best value than proposals AMD in efficiency with full load, but it at the same time it is inferior to them on the energy consumption in the state idle time. However, as far as the comparison of results athlon 64 X2 4800+ and processors, based on core windsor is concerned conclusions is completely ambiguous. The processor with 65 nm by core seems more advantageous from the relationship performance per whatt in comparison with the usual athlon 64 X2. But in this case Energy efficient athlon 64 X2 4600+ with the peak load teaches to demonstrate smaller energy consumption than CPU with 65 nm core.
It turns out that the efficiency of new processors AMD is not too revolutionary. A similar energy consumption AMD was possible to obtain also on processors with old core windsor.
Furthermore, the maximum frequency Of energy efficient for processors AMD, which are based on core windsor does not exceed 2.4 GHz. However, with the aid of the new core AMD obtained the possibility to increase frequencies CPU with the thermal packet 65 W to more imposing 2.6 GHz (but over the long term for 2.8 GHz).
Conclusions
The new core brisbane, produced on 65 nm technological process, will hardly cause revolution on processor market. From the view of usual user the new processors at basis of which will be used this core differ a little from their predecessors. In the new cores there are no micro-architectural improvements whatever, their frequency potential do not differed from the frequency potential of processors with the previous core windsor .
It is obvious that main dividends from the introduction Of brisbane will obtain AMD . The reduction in area of processor core together with increase in the diameter of the utilized in the production semiconductor plates will make possible to reduce the prime cost of dual core CPU. Furthermore, in the new core was the introduction of fractional coefficients multiplication witch will give company the possibility for more flexible processors. we should not forget also about the fact that after mastering 65 nm process amd will release more promising CPU with the optimized micro-architecture K8L, So AMD obtained the possibility to fix equipment and to attain good technological parameters to that moment
However, as far as the consumer qualities of new CPU, based on core Brisbane are concerned, here it is possible to note the positive motions. In the first turn, the processors with 65 nm core are more economical . All new models, based on core brisbane, will be unconditionally entered in the requirements of thermal packet 65 W. Secondly, from the new processors it is possible to expect somewhat lighter overclock.
xtreview is your : Video card - cpu - memory - Hard drive - power supply unit source
we would be happy to answer for your question . if you have suggestion or comment
regarding this review our support would be glad to help just join our forum and ask u will get the best answer
to discuss check our forum section :-)