Supercomputers prove to be very helpful in scientific experiments. A supercomputer is usually made of clusters upon clusters of smaller interconnected computers that work in parallel. If your current desktop PC has 2, 4 or 8 cores, supercomputers may combine more than 1,000 multi-core CPUs to provide enormous number-crunching power. However, there’s another possibility to constructing a super computer. Instead of using many CPUs, one may use the GPUs found in graphics cards. It is known that GPUs are superior to CPUs when it comes to complex calculations. That’s why Nvidia came up with CUDA, a technology that allows you to use up to 4 graphics cards that would provide roughly the same performance as a supercomputer cluster consisting of hundreds of PCs.

The ASTRA research group, part of the Vision Lab of the University of Antwerp, decided to extensively use the CUDA technology to develop new computational methods for tomography. In order to achieve this, they put together the FASTRA system containing 4 MSI 9800GX2 graphics cards (8 GPUs in total|) which are said to deliver the same performance as more than 300 Intel Core 2 Duo 2.4GHz processors. Of course you don’t see much improvement in games while using quad-SLI configurations, but that’s because the graphic drivers are not quite fully developed and the game engines are not optimized for such configurations. Otherwise, the CUDA does tomographic calculations very efficiently and consumes a lot less power than a supercomputer cluster.

Here is a list with the components used for the $4,000 FASTRA system:

• AMD Phenom 9850 processor + Scythe Infinity CPU cooler
• 4x MSI 9800GX2 graphics card
• 4x 2GB Corsair Twinx DDR2 PC6400 memory
• MSI K9A2 Platinum motherboard
• Samsung Spinpoint F1 750GB HDD
• ThermalTake Toughpower 1500W Modular PSU
• Lian-Li PC-P80 Armorsuit case
• Windows XP 64-bit

The biggest problem of the system is that these four dual-GPU graphics cards are cramped together and generate quite a lot of heat. The FASTRA uses aircooling and with the sidepanel removed the GPUs run at 55°C in idle, 86°C under full load and 100°C under full load with the shaders 20% overclocked. Hey guys, how about using some water cooling system if not liquid nitrogen? Also check this demo clip for more details:

This world is moving faster with each day, and when talking about computers, higher clock speeds require better cooling. One of these days, I was talking with a friend of mine about its next upgrade - he's going to get an Intel E8400 CPU, and intends to push it somewhere past 4GHz, on water cooling. We didn't talk about the video card he should use yet, especially since he doesn't have much time left for games, but I am sure the BFG 9800 GTX w/ ThermoIntelligence Copper water block would be a great choice for him, and maybe even for some of you, as we're going to see right away...


As you can see in the image above, this card looks great, and it won't take up too much space inside your case, either. The water block on the BFG NVIDIA GeForce 9800 GTX H2OC 512MB graphics card is a custom one co-developed by BFG and Danger Den, and you can bet they overclocked it to the highest stable level, but you can always try to go higher, of course!

"With the addition of our custom high-performance water block, we’ve been able to safely overclock the 9800 GTX to an unprecedented 780MHz on the core clock, and 1944MHz on the shader clock. Each card is hand-built by a trained BFG technician using Arctic Silver® thermal compound, and is exceptionally cool and totally silent—delivering up to a 28°C lower GPU operating temperature than reference cooling solutions," said John Malley, senior director of marketing for BFG Technologies.

Since I am sure you were wondering about it, here you have the frequencies:
- Core Clock: 780MHz (vs. 675MHz standard)
- Shader Clock: 1944MHz (vs. 1688MHz standard)
- Memory Data Rate: 2320MHz (vs. 2200MHz standard)

Unfortunately, the BFG NVIDIA GeForce 9800 GTX H2OC 512MB graphics card is available only in limited quantities at leading online retail outlets throughout North America and Europe, so you shouldn't think to much about it - grab it or miss it!

Getting cheap GPUs or CPUs and overclocking them afterwards, thanks to their huge overclocking potential, has been a nice budget trick for some time now. Most of the overcloking world records have been achieved using mid-range products and some manufacturers even started to encourage this sort of performance improvement by introducing special mechanisms that simplify the whole process. Even the GPU integrators like to show off some cool boosters on their cards, and to sustain this further, MSI comes along with a beefed-up GeForce N9600GT Diamond graphics card.

The new card was demonstrated at the ECOlution press conference in Taipei and the audience saw that MSI added a bright red overclock button to make overclocking a stroll in the park for non-experimented people who’d like to push their cards to the limits. When pressed, the graphics core and memory clock speeds are overclocked, and the voltage increased.

To further increase the overclocking potential, MSI added 2GB of GDDR3 memory to the Diamond version which also comes with a whole bunch of connectors: HDMI, DisplayPort, dual DVI and optical audio output. The card will also be bundled with the Dr.MOS software that monitors real-time performance and cooler speeds.

Although the card is going to be released in June, there still are some minor problems to fix. In SLI dual graphics mode the button won’t work, but MSI are keen to point out that it’s definitely on their to-do list for the near future. Moreover, we don’t know yet how far the overclocking capabilities will go. The card is expected to cost around $250.

Despite the fact ATI is still far from having a clear lead in the discrete graphics market (they don't even hold the first position, anyway), they always managed to arrive first in some areas - first cards using GDDR4, then first cards with DirectX 10.1 hardware support, and now even the first cards with GDDR5 support! The HD 3xxx generation of video cards is still pretty fresh, but it seems we already have the specifications of the upcoming goodies, the HD 4870 and HD 4850!


Regarding the standard dimensions, it's enough to say both cards will have the same length, and that is 23 cm(9 inch). UVD 2 promises to allow hardware decoding for Blu-Ray 1080p content, and even beyond, while the on-chip HDCP feature allows easy playback of HDCP protected content. At last, there's the HDMI support we're already used to, but wait and see the real goodies, because we're just getting started...

The memory to be used with these cards is going to be GDDR3/GDDR5, probably depending on the manufacturer of a certain model. The DirectX support is still at 10.1, probably waiting for NVIDIA to catch up(;)). The 256-bit memory interface isn't something out of this world, but dynamic geometry acceleration and game physics processing sound really great, together with the 24X custom filter anti-aliasing.

No, I have no idea about pricing yet, but rumors say that the R700-based cards, including the HD 4870 and HD 4850, should arrive in August. If this proves to be true or just a rumor, that's what remains to be seen, but you'd better start saving now, because I have the feeling we're going to have a 9800GTX killer from ATI for less than $250!!!

Nvidia has finally revealed the official specs for the GTX 280 and GTX 260 GPUs at its famous editors' day. Te specs are pretty close to what rumors were announcing previously this month. Let’s take a look at the real deal.

GTX 280 has the core clocked at 602MHz, while its 1GB of GDDR3 is complemented by a 512-bit memory interface and clocked at 1,107MHz. As a high-end GPU, GTX 280 integrates 240 stream processors and 32 ROPs with Shader units running at 1,296MHz. A graphics card equipped with this GPU has an estimated TDP of 236 Watts, and the recommended price is around $600+ at launch.

The more affordable GTX 260 is clocked at 576MHz , and if we compare it with the GTX 280 clock, we see it’s not a big difference after all.. However, the GTX 260 will have its GDDR RAM working with an awkward 448-bit memory interface, incorporating 896MB of GDDR3 memory clocked at 896MHz. It will also include only 192 stream processors and 28 ROPs with Shader units clocked at 999MHz. The TDP factor is estimated at 182 Watts and the launch price tag should be $449.

ATI might still have a chance with their RV770Pro chips. They won’t be able to compete with Nvidia’s models performance-wise, but they sure have the upper ground with reduced TDP factors and more affordable prices. The performance war is set to start sometime in August when ATI’s dual-GPU cards are expected to launch.