by Ilya Gavrichenkov
11/29/2006 | 09:01 PM
When Intel introduced their new Core micro-architecture, it ensured a significant performance improvement of Intel processors. As a result, AMD had to yield the technology leadership to the competitor. Actually these days AMD solutions remain acute only in the lower price segments, offering single-core processors, which Intel hasn’t yet transitioned to the new micro-architecture. Nevertheless, AMD keeps fighting bravely against Intel in all market segments (although with not much success yet). Their major task in this fight is to retain the army of dedicated fans until the arrival of the new K8L architecture that the company pins a lot of hopes on. And in the meanwhile they resort to marketing and price tools.
One of those marketing solutions destined to help AMD withstand Intel’s success is the launch of AMD Quad FX platform (also initially known as 4x4). This platform is officially launching today and should become AMD’s response to the recent quad-core Intel Kentsfield launch. Unfortunately, AMD is not yet ready to release any fully-fledged CPUs with more than 2 cores. There are architectural reasons (the use of system memory to exchange data between the processor cores) as well as technological reasons (no mass production technology) for that. Therefore, until they finish the development of new micro-architecture that will know to organize cores interaction without involving the memory bus and until they introduce the new 65nm production technology that will give out smaller processor dies, AMD has to look for some other ways to compete with Intel Kentsfield.
In fact, AMD Quad FX is a dual-socket system targeted for high-end desktop applications. This platform is designed to accommodate a pair of dual-core processors so that there are four computational threads used simultaneously. In other words, AMD Quad FX can be regarded as an alternative to Intel Kentsfield from this prospective, especially considering a system as a “black box”. In addition AMD is going to price the components of this platform accordingly, so that AMD Quad FX system could belong to the same price range as systems built around Intel quad-core CPUs.
Note that Quad FX doesn’t really strike us with its intellectual design when we cast the very first glance at the CPUs that AMD intends on using in this platform. And here is why we think so. The current K8 micro-architecture AMD is using these days allows assembling multi-processor configurations very easily. One of the things that make it easy is the three HyperTransport busses in AMD processor cores. Nevertheless, the Socket AM2 infrastructure that appeared in the market about half a year ago turned out absolutely unprepared for the multi-processor configurations. This 940-pin processor socket has contacts for only one HyperTransport bus, which is surely not enough even for a dual-CPU configuration. Therefore, only a few months down the road from the Socket AM2 desktop processors launch, AMD had to start promoting a new socket for desktop Quad FX platforms.
Since there was not that much time available, AMD didn’t design an absolutely new processor socket this time. For the new Quad FX platform concept they decided to go with the 1207-pin Socket F that is moreover designed in LGA form-factor. Sockets like that are used in contemporary server systems for AMD Opteron CPUs and Registered DDR2 SDRAM.
I have to admit that the use of Registered DDR2 memory wouldn’t be the best decision AMD could make for its desktop platforms. Although Registered memory is much more stable, it is also more expensive than the regular DDR2 SDRAM and features higher latencies. Therefore, AMD decided to replace the memory interface in the Quad FX systems: these CPUs use regular Unbuffered DDR2 SDRAM. As a result, Quad FX systems also lost their connection with the server infrastructure.
Taking into account everything I have just mentioned above, AMD had to create a few absolutely new dual-core processors that are not compatible with any platform other than Quad FX. They were formally placed within the Athlon FX processor family. Just like the today’s Athlon 64 FX for Socket AM2, these processors are based on 90nm Windsor core, however they boast three active HyperTrasnport connections, one of which serves to connect the CPUs with one another.
Today AMD is offering three models of dual-core Socket 1207 processors that are compatible with the Quad FX platform. They are Athlon 64 FX-70 working at 2.6GHz clock speed, Athlon 64 FX-72 working at 2.8GHz and Athlon 64 FX-74 working at 3.0GHz. They will launch one more faster processor in Q2 2007: AMD Athlon 64 FX-76 with 3.2GHz clock speed. All these CPUs feature 2MB total L2 cache memory shared between the cores. The detailed CPU specifications for the AMD Quad FX platform look as follows:
And this is the currently available product line-up:
Athlon 64 FX-74
Athlon 64 FX-72
Athlon 64 FX-70
HyperTransport, 2 GHz
1.35 – 1.4 В
1MB + 1MB
90 nm, SOI
Price for a pair
Note that the above listed CPUs are sold only in pairs, so you cannot purchase a system with only one dual-core CPU intending to upgrade in the future.
Two Athlon 64 FX-74 processors we got in our lab
As we see, AMD is trying to create analogies between its new Quad FX platform and the quad-core competitor’s solutions. However, from the compatibility standpoint the new AMD platform looks less attractive than Kentsfield. It can be called “a thing in itself”: unfortunately, it is not compatible with any other platforms and requires only special dual-core CPUs.
However, AMD do their best to present this fact from an advantageous standpoint. They claim that Quad FX will not be just a temporary solution temporarily making up for the absence of quad-core processors in the company’s product range. AMD promises this platform will live much longer. Moreover, when the quad-core processors also known as Agena (former Altair) are ready, they will be also adopted by the Quad FX platform. In other words, AMD is introducing Quad FX as a new innovative platform for computer enthusiasts who want to get the maximum performance at any price. Today this platform provides four simultaneous computational threads, but in the second half of 2007 it may be enhanced to be able to process up to 8 parallel computational threads.
By setting special prices for the CPU kits AMD managed to get the total cost of the new Quad FX platform very close to that of the Intel Kentsfield based systems. However, from the today’s popular performance-per-watt viewpoint, the situation is not that rosy at all. Athlon 64 FX processors for Quad FX platform manufactured with 90nm technology feature typical heat dissipation of 125W each. As a result, two CPUs like that will consume almost twice as much power as the top of the line quad-core Intel processor. So, from the efficiency standpoint, the new AMD Quad FX platform is losing to the Intel competitor big time. As for the performance leadership, our test session will reveal, who the winner is shortly.
Note that since AMD Quad FX is a fully-fledged dual-socket system it features two dual-channel DDR2 SDRAM controllers. As a result it theoretically features twice the memory bus bandwidth of the regular desktop platform and hence allows resolving some of the problems during data transfers between the cores. The communication between the processors and the memory is implemented in the NUMA technology (Non-Uniform Memory Architecture) that allows both processors to operate within the same address space. In other words, each processor can address the memory of the other processor via the HyperTransport bus between them. And no special tricks are required in this case. Moreover, it doesn’t really matter for either of the CPUs in which memory the requested data is located. However, it is important to understand that if this data is located in the memory of the other CPU, the latency of the corresponding operation will be significantly higher than in case the data were in the processor’s own memory.
Quad FX platforms can use the shared memory in two ways, which is controlled from the BIOS Setup by adjusting Node Interleave option accordingly. With Node Interleave enabled (the nodes in this case are the processors and the memory controller with the DDR2 SDRAM modules connected to it) the memory fills up evenly no matter what CPU initiates the writes. In this case the data appears “spread’ over both memory subsystems of the two CPUs. Of course, it results into higher latency of memory operations than in case of regular single-processor systems, because half of all the memory requests go through the memory controller of the “other” CPU.
With Node Interleave disabled, the processor writing data into the memory first of all uses its one controller. However, this approach also doesn’t allow Quad FX systems to reach the efficiency of the single-processor platforms. The thing is that Microsoft Windows XP operating system does its best to load all CPUs of the multi-processor system evenly and hence constantly switches the tasks between the processors. As a result almost in half the cases the CPU will still have to address the memory subsystem of the other processor with higher latency.
As a result, it is important to understand that from the memory performance standpoint, single-processor AMD systems are overall faster than multi-processor ones. The launch of Microsoft Windows Vista should with enhanced scheduler should make things better for the Quad FX platform, because this scheduler will support NUMA technology and will not shift tasks forth and back all the time between the logical processors.
That is why AMD recommends enabling Node Interleave for Windows XP and disabling this feature for Windows Vista on its Quad FX platform.
AMD Quad FX platform required not only new CPUs, but also new chipsets and new mainboards. AMD selected Nvidia to be its chipset partner for a reason: they have quite a bit of experience in the server as well as desktop chipset development. Especially, since Nvidia has long had all the necessary components for the solution AMD needs for its Quad FX system.
For the new Quad FX platform Nvidia introduced a special core logic set called nForce 680a SLI. The name of this product sounds very similar to nForce 680i SLI we have reviewed recently (for details see our article called Nvidia nForce 680i SLI Chipset Review) and there is a reason for that too. Just like the new chipset for the Conroe processors, Nvidia nForce 680a SLI contains fifth generation nForce component chips. To be more exact, nForce 680a SLI is none other but a combination of two nForce 570 SLI and there is nothing really innovative about it. Actually, you can even see from the chart below that nForce 680a SLI = 2 x nForce 570 SLI.
Nvidia nForce 680a SLI
Nvidia nForce 570 SLI
No doubt, Nvidia developed a special chipset for Quad FX without any effort. Especially since the two chips of this core logic set are connected with one another via one of the system CPUs. In fact, this should guarantee that Quad FX system will also run normally with only one CPU installed, not to mention that the overall system design will be simpler in this case.
No wonder that the diagnostic utilities still see the old chipset there:
As a result, Nvidia has every right to be proud of the fantastic features their nForce 680a SLI boasts, as it actually has twice the features of the nForce 570 SLI. This chipset supports two PCI Express x16 graphics busses and two PCI Express x8 busses. They can be used for SLI and Quad SLI configurations or for physics accelerators. Also there are a few PCI and PCI Express x1 busses available. Nvidia nForce 680a SLI supports 12 Serial ATA-300 ports, 4 Gigabit Ethernet ports and 20 USB 2.0 ports.
The only mainboard for AMD Quad FX platform that is currently available in the market is made by ASUS. It is called L1N64-SLI WS. Other mainboard manufacturers seem to be not yet interested in the new Quad FX platform. Which is actually not surprising at all, keeping in mind that the sales of this platform will not reach any significant numbers soon enough.
Now that we took a closer look at Nvidia nForce 680a SLI chipset, ASUS L1N64-SLI WS mainboard doesn’t surprise us with its remarkable features anymore. Its features are mostly those of the chipset. The mainboard is equipped with two 1207-pin LGA sockets with two DDR2 DIMM slots assigned to each of them. To ensure maximum performance AMD recommends using all four memory slots for two dual-channel memory kits. As for the coolers retention, it is pretty standard, so you can use traditional Socket AM2 cooling systems on ASUS L1N64-SLI WS mainboard.
Four physical PCI Express x16 slots laid out on this mainboard are color coded. Blue slots are logically connected to the fully fledged PCI Express x16 bus, while the black slots are actually none other but the PCI Express x8 bus. Besides the PCI Express x16 slots there is also one PCI Express x1 slot and one regular PCI slot on the PCB, which actually has hardly fit onto the mainboard. Note that ASUS L1N64-SLI WS cannot accommodate four graphics cards with dual-clot cooling system. The three graphics cards like that will block all other available expansion slots.
According to the Nvidia nForce 680a SLI chipset specification, the mainboard has 12 Serial ATA-300 ports laid out and even one Parallel ATA-133 port. The mainboard allows creating RAID 0, 1, 0+1 and 5 arrays. Each array like that can consist of maximum 6 hard disk drives connected to the same core logic chip. Also ASUS equipped its mainboard with one SATA On-the-Go port laid out on the mainboard back panel. There is a special external SiI3531 controller responsible for this port’s operation.
Out of 20 USB 2.0 ports supported by the chipset, ASUS L1N64-SLI WS features only 10. Four ports are placed on the back panel and the remaining six are available in the form of onboard pin-connectors. The mainboard also doesn’t make use of all the supported network controllers. ASUS L1N64-SLI WS offers only two Gigabit network controllers and their RJ45 connectors are laid out on the mainboard back panel.
To implement the FireWire interface, ASUS equipped its mainboard with VIA VT6308P chip. This controller provides two connectors for add-on brackets with IEEE1394a ports.
High-definition sound solution on ASUS L1N64-SLI WS uses AD1988B codec. The mainboard features six analog audio Ins/Outs, an optical and coaxial SPDIF Outs located on the back panel.
Also there are PS/2 ports for keyboard and mouse and a parallel port. The serial port is also available, but it is implemented as a pin-connector.
Since ASUS L1N64-SLI WS mainboard features two processor sockets and a lot of additional connectors it is designed in eATX form-factor measuring 307x267mm. As a result, far not all the system cases may be fit for the new Quad FX platform. And certainly not all the power supply units, which was clear from the very beginning. By the way, there are three power connectors on this board: 24-pin, 8-pin 12V one and a standard MOLEX.
Despite all ASUS engineers’ efforts, the design of their L1N64-SLI WS is far from perfect. The 24-pin ATX power connector is located in an unbelievable spot: almost right in the center of the PCB.
Each of the CPUs features its own four-phase voltage regulator. The MOSFET used in these regulators are covered with two copper heatsinks which can be topped with individual turbo-fans. The chipset is also cooled down with a very advanced solution. Both chips of the core logic set used on this board are located next to one another, so they share a cooler. The construction of this cooler includes a flat copper plate with two heatpipes leading to a separate copper heatsink also equipped with a turbo-fan. Despite this complex cooling system, the chipset of ASUS L1N64-SLI WS heats up so much, that you can burn your fingers if you touch the chipset heatsink during operation.
Since AMD Quad FX platform and ASUS L1N64-SLI WS are targeted for computer enthusiasts, the BIOS Setup of this platform features quite a list of advanced options, including overclocking-friendly ones, of course.
You can configure the memory subsystem, set all the major parameters including the important for AMD systems Node Interleaving option.
The mainboard also allows performing some processor overclocking. I cannot call the overclocking friendly features of ASUS L1N64-SLI WS unprecedented, but they are quite enough to achieve noticeable results. You can raise the clock generator frequency up to 400MHz, the processor Vcore can be increased by maximum 0.2V above the nominal, while the Vmem can be raised to maximum 2.5V.
I have to admit that we didn’t perform any detailed testing of the overclocking-friendly features of ASUS L1N64-SLI WS mainboard. But we checked that the mainboard works stable at 250-260MHz clock generator frequency. In fact, Quad FX processors hardly require any further overclocking. Firstly, their clock frequency multiplier is unlocked like by any other Athlon 64 FX, and secondly, their nominal frequencies are already very close to the maximum the Windsor core can handle thanks to the current production technology.
The main goal of our test session was to check whether AMD Quad FX platform can be regarded as an opponent to Intel quad-core processors for the performance prospective. For our tests we put together a few platforms using the following hardware components:
We tested with the mainboard BIOS settings adjusted for maximum performance.
Also note that following the manufacturers recommendations we used four 512MB memory modules in the Quad FX system, while all other testbeds were equipped with two 1GB DDR2 SDRAM modules.
SYSMark 2004 SE test measuring the system performance in real applications turned into a true awakening for those AMD fans who despite common sense have hopes that the new platform would prove a worthy competitor to Intel’s solutions until AMD comes out with their K8L architecture. The new AMD Quad FX platform loses not only to the quad-core Kentsfield processors, but even to the dual-core Conroe processors.
If you take a closer look at the diagrams, you will be able to notice another pretty strange thing: Quad FX platform based on two dual-core Athlon 64 FX working at 3.0GHz loses even to one dual-core Athlon 64 FX working at 2.8GHz. However, there is nothing unnatural about it: the reason lies with the organization of the memory subsystem in the AMD dual-socket platform. Although each CPU has its own memory subsystem, the data gets distributed evenly over all the memory banks independent of the CPU they are assigned to. Therefore, each processor has to address not only its own memory subsystem, but also the memory subsystem of the other CPU. These “long-distance” requests feature much higher latency and hence slow down the system. And if we add that the data transfer between the CPUs is performed solely through the memory subsystem, then the performance drop we observe in dual-socket Quad FX platform will be absolutely logical.
As a result, the applications working actively with the memory subsystem may turn into a so-called Achilles’ heel for Quad FX platform.
Another test, PCMark05, produced better results on Quad FX. At least, this dual-processor platform with dual-core CPUs in it outperformed the single-processor system with one Athlon 64 FX-62. Although, working at 3GHz speed it cannot catch up with the quad-core CPUs from Intel. Looks like contemporary AMD processors cannot compete against Intel Core micro-architecture even within a dual-processor system.
Nothing new appears in Futuremark 3DMark06 diagrams. Quad-core Intel processors, including Core 2 Quad Q6600 working at 2.4GHz clock frequency, turned out faster than the Quad FX platform with two 3GHz Athlon 64 FX-74 processors.
The results of the processor test are better. However, the high-performance AMD platform can only beat Core 2 Quad Q6600 at the most, and is still behind the quad-core Core 2 Extreme QX6700.
The situation in real games is again quite sad. Although AMD positions this platform as a high-end solution for gaming enthusiasts, its performance is lower than that of a system built around a single Socket AM2 Athlon 64 FX-62. This result is once again explained by the high latency of the memory subsystem supporting NUMA technology, even though contemporary games do not need to process four computational threads simultaneously. So, this is a great example that the effect of server technologies transferred into the desktop segment is not always positive.
However, I have to stress that Quad FX may become very fit for some games in the near future. However, in this case we will certainly need the games that can efficiently load all four cores of the system simultaneously. At this time there are no applications like that in the market. However, there is hope: AMD’s press presentation suggests that in 2007 there should be more than 20 gaming titles already.
Well, looks like it doesn’t make much sense to continue investigating the gaming performance of the AMD Quad FX platform at this time. Let’s just take a look at the performance of this platform in multi-threaded strategy games, like computer chess, for instance:
The situation here is a way better than what we have seen before in 3D shooters. Dual-socket Quad FX platform ensures 90% performance improvement compared with the single Athlon 64 FX-62 (for processors running at the same clock speeds). No wonder: chess algorithms do not require a lot of RAM, but they can be split into parallel threads very easily. However, Intel’s quad-core processors on Core micro-architecture outperform Quad FX in this case.
Quad FX platform performs quite well during video encoding. At least the codecs supporting multi-processor configurations work faster in this system than in a single-processor testbed with Athlon 64 FX-62. However, once again the ultimate leaders are Intel’s quad-core CPUs.
Quad FX platform performed as fast as a system equipped with one dual-core CPU in the codecs supporting only two threads maximum. And it again means that it is behind Intel processors on Core micro-architecture.
Adobe Photoshop and Adobe Premiere Pro can very well load a dual-processor system with dual-core processors. Therefore Quad FX platform works faster than a regular Socket AM2 system here. However once we compare its performance against that of Intel Kentsfield, the results will be far not in favor of our hero, especially in Photoshop.
Quad FX platform loses to Core 2 Extreme QX6700 during final rendering in different applications.
The results in viewports are very similar to what we have just seen in gaming applications. Quad FX works even slower than the Socket AM2 platform with a single dual-core Athlon 64 FX-62.
In this section we performed a few tests aimed at revealing the systems performance when there are a few applications running simultaneously. We launched several resource-hungry tasks in different applications and measured the time it took the test platforms to complete them.
Here we were processing an image in Adobe Photoshop and at the same time compressing a folder with files using WinRAR utility. The results are quite common already: Quad FX platform is faster than any dual-core processors, but loses to Intel’s quad-core CPUs on Core micro-architecture.
Then we were editing video in Adobe Premiere Pro and encoding an MP3 file using Apple iTunes. And here, finally, AMD’s dual-processor system with two 3.0GHz CPUs managed to outperform a quad-core Intel processor working at 2.4GHz.
The third test is probably the most sophisticated. Our test platforms were busy working on three tasks at a time: image editing in Adobe Photoshop, final rendering in 3ds max and video encoding into MPEG4 format. Here Quad FX platform performed not so well falling behind Intel’s dual-core CPUs. Moreover, as we see, the dependence of Quad FX performance on the CPU frequency is quite small, so I would assume that there is something limiting this performance dramatically. But of course, it is high latency of the memory subsystem, which we have already complained about several times.
Besides the performance of the dual-core and quad-core platforms running several parallel tasks, we also decided to investigate the influence of background processes on the performance in certain resource-hungry applications. We measured the fps rate in Quake 4 game with several copies of WinRAR utility running in the background.
We have finally found the strength of the Quad FX platform that actually originates exactly from the server side. We can see that as the number of simultaneously processed tasks increases, Quad FX platform starts outpacing Intel based systems. And this is quite logical, actually. Sooner or later Intel processors use up the entire bandwidth of their memory subsystem that serves for data transfer between the Kentsfield core pairs that have no shared L2 cache. Quad FX platform features two independent memory controllers in two CPUs and hence features twice as high memory bandwidth. Therefore, Quad FX turns out more efficient than a much faster quad-core Intel Core 2 Extreme QX6700 once the multi-threaded workload increases dramatically.
Although we have to admit anyone will rarely have five or more resource-hungry applications running at the same time. Therefore, this advantage of the AMD Quad FX platform can hardly be determinative for common users.
Well, our tests revealed that AMD Quad FX platform cannot boast undefeated performance. Maybe it could be corrected by successful overclocking? Turned out it couldn’t. Our experiments showed that it is extremely hard and hardly efficient to overclock the Quad FX platform.
As we have already mentioned above, ASUS L1N64-SLI WS mainboard Quad FX platform is built on has quite a few overclocking friendly options for processor tweaking. Moreover, it allows increasing the clock generator frequency above the nominal value and adjusting the processor clock frequency multiplier that is unlocked by default by all Athlon 64 FX processors. We can even confirm that ASUS L1N64-SLI WS runs absolutely stably at 250MHz clock generator frequency. However, when it came to practical side of the Quad FX overclocking, we faced some issues.
Our test platform used two Athlon 64 FX-74 processors with 3.0GHz nominal clock speed. Unfortunately, we couldn’t increase this frequency any further. If the clock generator frequency was set even 1MHz above the default 200MHz, the system wouldn’t boot Windows XP shutting down automatically. It was pretty strange also because nothing could be done to resolve this situation: no common tricks used to improve the stability of the overclocked system, such as increasing the processor voltage, bus voltage or chipset voltage or lowering the HyperTransport bus multiplier, would help. We could free some room for the clock generator frequency increase by reducing the processor clock frequency multiplier below the nominal value, but only until the CPU frequency hit 3.0GHz. Once the frequency got higher, the situation repeated: the system would simply shut down on us.
Everything we have just said suggests that AMD, Nvidia or ASUS introduced some kind of protection into their system that prevents the CPU frequency from being increased above the nominal. Of course, it is really hard to believe especially since AMD has never prevented overclockers from having their fun, but it is an undeniable fact. It is even harder to believe that the CPU is working at the utmost of its potential in the nominal mode already.
Unfortunately, this is not all yet. Another huge problem of the Quad FX platform is its extremely high power consumption and heat dissipation. It is simply impossible even to compare it against anything else from this prospective. Of course a system built with two 90nm processors working at their maximum speed will anyway eat up more power than the competitor’s solutions built with only one CPU (even a quad-core one) manufactured with finer 65nm process. Besides, the Nvidia nForce 680a SLI chipset also contributes a lot here, as it consumes much more power than any of the Intel chipsets.
To back up what we have just said, we measured the power consumption of the Quad FX platform (in the tested configuration) with two Athlon 64 FX-74 processors. Even in idle mode with enabled Cool’n’Quiet technology (without the monitor power consumption of course) this system consumed 192W of power, which is more than most contemporary systems would require for 100% workload. As for the maximum power consumption level, we managed to hit 501W when the system was running fully loaded. However, it is no secret that Quad FX systems can consume much more than that, because this time we didn’t take into account the power consumption of the video subsystem, as it wasn’t loaded at all. In other words, you will have to get a special powerful PSU for your Quad FX system, that’s for sure.
Another problem directly connected with the power consumption is the level of generated noise. All this heat from the system should be dissipated somehow and this is when we often use traditional air-coolers. Without the graphics card Quad FX platform already has 5 powerful fans cooling the processors, the chipset and voltage regulators on the mainboard. Of course, there are at least a few additional case fans. And the coolers of the graphics cards and system PSU. So, there will be quite a few rotating parts in Quad FX platform for sure, so it should never be even considered quiet in the first place.
Here I would also like to add that you should make sure you’ve got the right case to accommodate your Quad FX system, far not any case will do the trick, because ASUS L1N64-SLI WS mainboard is relatively big and the system case should ensure proper cooling of all the components.
AMD Quad FX platform turned out far from what we have expected. Although this way AMD intended to respond to Intel’s quad-core processors launch, this response turned out quite strange I should say.
First of all, the performance of a dual-processor platform built with two dual-core Athlon FX processors turned out lower than that of the competitor’s solutions built on quad-core Kentsfield CPUs. We have seen this in all test applications throughout the entire session.
Secondly, Quad FX platform is often slower than the regular Socket AM2 system with a single CPU because of the higher memory subsystem latency. NUMA technology that proved highly efficient in servers turned out to do more harm than good in the desktop space.
Thirdly, from the performance-per-watt prospective Quad FX platform loses not only to Intel Kentsfield based solutions but to all other platforms as well. The sky-high heat-dissipation and power consumption of this platform also set specific requirements to power supply units, system cases and system cooling.
In other words, AMD Quad FX will most likely appeal only to the most dedicated AMD fans. For everyone else this platform will most likely be none other but a unique desktop system prototype showing the possibility (or maybe even uselessness) of introducing AMD’s server technologies into the desktop sector. Although we cannot deny that the upcoming AMD processors based on promising K8L micro-architecture will give new meaning to the exciting Quad FX concept.
True, all the drawbacks we have listed in relation to the AMD Quad FX platform in this article have primarily to do with the Athlon 64 FX processors. As for the platform as a new concept, then it shouldn’t be underestimated. Of course, the today’s AMD processors based on old K8 micro-architecture and manufactured with 90nm production process have hard times competing with Intel’s solution built using the newest Core micro-architecture. Therefore, once AMD refreshes its Athlon 64 FX product family by switching to finer 65nm process and introducing promising K8L micro-architecture, Quad FX platform may acquire new meaning. In fact, we have nothing against the dual-processor desktop concept. Especially since the drawbacks connected with the incorrectly working NUMA technology will most likely disappear next year when we all transition to new Windows Vista OS. Therefore, as soon as AMD provides the market with new CPUs boasting competitive power consumption and heat dissipation levels. Quad FX platform will certainly become an excellent high-performance solution for wealthy computer enthusiasts. Well, let’s just wait: our patience should be rewarded.