IDF Spring 2006 Coverage: Day 2

Today I am going to share some details on Intel’s own advanced liquid cooling system an, of course, the Intel Conroe performance details in some popular synthetic and gaming benchmarks.

by Anna Filatova
03/09/2006 | 05:47 PM

Intel Advanced Liquid Cooling Technology

In his yesterday’s presentation Gavin Stanley, Intel Thermal Mechanical Engineer, talked about some interesting cooling technology that they were developing within the company. In their intention to meet the needs of extreme users they realized that it is not enough to simply offer a high-speed processor. They also need to make sure that the users can really enjoy the potential of the product and that it is reliable enough.

 

The reliable and stable performance in extreme user environment can only be obtained with proper cooling, which is easy to mount, inexpensive and reliable throughout the entire service time. This is what Intel advanced liquid cooling solution can provide the users. To ensure reliability of operation each product uniquely and individually tested.

This cooling solution uses integrated and modular pump and cold plate approach. Let’s take a closer look at it.

As we all know, there are 4 major components in a traditional liquid cooling system:

In Intel’s case the separate components and associated tubing have been removed thus allowing to eliminate transitional and frictional losses in the system. The pump is literally wrapped around the outside of the circular cold plate. DC brushless motor is integrated into the pump directly.

This approach also has some cost advantages: basically there is one unit that fulfills the heat transfer and fluid movement functions at the same time.

Here Intel specifically stressed that although you may get this impression at first, they were not after and not seeking exotic technologies and industry solution. From the implementation standpoint the hydrodynamic bearing technology allows to reduce cost.

From the reliability standpoint Intel offers very robust motor technology. Moreover, every unit is leak and seal tested.

As you may have noticed from the picture above, the tubing is firm. The reason for this material choice is the fact that PCV based or rubber based tubes cause the majority of the fluid loss in the system. If you decide to go with a hard tube approach and implement it properly, you may not need a separate reservoir unit, because the fluid loss throughout the entire service time will be not so great and you do not need an extra reservoir for reserve purposes.

After the system is assembled, they fill it with liquid through the charge port. The liquid inside is 35% propylene glycol aqueous solution. After that the charge port is sealed for life (white top). Then the unit undergoes all the necessary tests to ensure that there are no leaks or pressure losses.

According to Intel, this cooling solution is much more efficient than many others out there thanks to the construction peculiarities. Namely, the fluid is entering the module through the center of the pump. It is impinging into it, and it allows high heat transfer coefficient than in case of parallel contact. Besides, we have already mentioned that the pump is kind of wrapped around the cold-plate, and by doing that we get a larger radius of the cooling module, which also affects the efficiency positively.

Intel advanced liquid cooling solution can fit into ATX as well BTX infrastructure. Here it is:


ATX system on the left, BTX system on the right

On the show floor we could see a gaming system equipped with the new Intel advanced liquid cooling solution that was running at an unprecedented speed of 5GHz in Quake 4 game. Here is a screenshot:

Conroe Performance Unveiled

Now comes the most exciting part I assume. X-bit labs got a great opportunity to take a closer look at the new system based on the upcoming Conroe processor running side-by-side with a competitor solution. As you may have already heard from some other sources that were among the lucky ones to play with these systems, the upcoming Intel Conroe processor demonstrated unprecedented results sometimes exceeding those of the competitor by the good 30% or even more. But before we take a look at the results that I managed to obtain on these two systems, I would like to share some thoughts with you.

I was given the opportunity to run a few synthetic (PCMark05) and gaming (Quake4, F.E.A.R. Half-Life 2) tests on two platforms. The Intel platform was assembled with the Conroe processor working at 2.6GHz, Intel D975XBX BadAxe mainboard based on Intel’s 957X chipset. As for the competitor platform it was an AMD Athlon 64 FX-60 overclocked to 2.8GHz frequency running in DFI LanParty UT RDX200 mainboard on ATI RD480 chipset. Both systems were equipped with two ATI Radeon X1900 XT graphics cards in CrossFire mode and 1GB of memory. The Intel platform supported DDR2, of course, while the AMD platform was still running with DDR SDRAM configured with very aggressive timings of 2-2-2 and 1T Command Rate.

This is where I feel like pointing out my first big concern. We are comparing two different platforms here: DDR1 against DDR2. Even though Intel claims that AMD should benefit not more that 5-10% by switching to DDR2 this is not a very convincing argument, IMHO. The implementation of the memory controller may have very significant influence on the performance, so I wouldn’t be so certain about the 5-10% performance boost.

The second concern I would like to share with you is the fact that we are actually comparing the current AMD processor against the Intel solution, which will be current in at least three months. Three months is a lot of time. In three months AMD has every chance to roll out something a way faster than their current FX-60, don’t you think so? So, is this a fair comparison? Intel claims this is the best approximation of what the high-end competitor to the upcoming Conroe platform is going to be. Well, it is not really fair, IMHO, but it certainly has great marketing power, that’s for sure.

My third concern is that at this point I didn’t have any X-bit’s original demos at hand, so I had to use the timedemos integrated into all the gaming applications. And you know the deal with timedemos: people get very suspicious about the results, because the temptation to add a little optimization here or there is too big. Here we can only bet on Intel’s word that no optimizations have been made whatsoever.

Now that I shared with you some of my thoughts that will hopefully prevent you from making any hasty conclusions, we can move on to the actual benchmark results.

F.E.A.R.: The first set of tests were run with graphics quality set to Low in order to eliminate the influence of the graphics subsystem on the results and get a more illustrative CPU performance.

Intel Conroe

AMD Athlon 64 FX-60

The second set of tests were run with the Maximum graphics quality setting (1280x960 resolution):

Intel Conroe

AMD Athlon 64 FX-60

Quake 4 (timedemo): the results in Quake 4 suggest that Conroe is going to become a CPU of choice for gamers. Here are the results:

Intel Conroe

AMD Athlon 64 FX-60

Half-Life 2: Lost Coast (timedemo): the demo was run two times on each system with the 1024x768 resolution and Advanced settings. The performance advantage of the Conroe system against the competitor was over 30%.

Intel Conroe

AMD Athlon 64 FX-60

PCMark06 : The CPU test set from PCMark provided the following results:

Intel Conroe

AMD Athlon 64 FX-60

Well, the results are very impressive. However, I would refrain from conclusions for the reasons mentioned above. Hopefully very soon we will get more opportunities to test this platform and perform a truly fair contest between the two eternal rivals.