A yearly tradition for us at Hardware Upgrade is visiting CeBIT in Hanover,Germany, the largest exclusive computer and IT expo in the world. CeBIT is held every spring on the Hanover fairground. The large exhibition is quite imposing and its sheer size makes it very difficult to visit all of the company exhibits during the time allotted for the show. Add this to a massive amount of meetings and press conferences, and it is easy to see how CeBIT becomes an exhaustive time of the year for us.

Hannover Fairground: map of the CeBIT 2007 showroom floor

For us CeBIT can either be packed with exciting new products and technologies, or it may end up being rather boring with only rehashed technologies and products on display. This year, from our point of view it unfortunately ended up being the latter. The surface area of the expo was reduced this year, and the length of the tradeshow was reduced from 7 days to only 6 days, a clear indication that CeBIT is not gaining as much attention as it used to in years past. CeBIT is not the only tradeshow that is receiving less attention. For the most part, nearly all technology tradeshows have been suffering in the past few years, primarily because large manufacturers are beginning to change their perspective on the benefits of tradeshows. Tradeshows are being used less and less to introduce new products, while private launch parties have really begun gaining popularity because it allows both the press and principal clients to be present.

The World Wide Web has also contributed to the decline of the traditional tradeshow. Many of the largest manufacturers give nearly complete coverage of their products in realtime using webcasts, meaning the expense of a large exhibit at a tradeshow is starting to look much less justifiable. This also means that many journalists and individuals don’t need to travel to Hanover to see new products and can instead directly view them from the companies’ webcasts. In addition, the overall aim of CeBIT has become a tradeshow to conduct business and meet potential clients and less of a tradeshow targeted at technology enthusiasts

Cebit 2007, therefore, turned out to be a rather dull experience for us, especially in terms of video cards. The postponement of new DirectX 10 cards from NVIDIA and AMD limited the amount of new cards that we saw at the show, and for a few exceptions the cards were mostly things we had already seen. This report will cover the video cards of CeBIT. Future reports will cover other aspects of the tradeshow, including memory, motherboards, and chipsets.

ATI’s new upper end video cards are definitely on the minds of many hardware enthusiasts; however, they are even more on the minds of ATI-AMD’s manufacturing partners. The launch of these video cards, as is already widely known, has been in fact postponed a number of times. In the latest postponement, the launch was actually pushed back from March all the way to the second quarter of 2007. ATI plans to release these products to the press in the second half of April, therefore indicating that the debut of the new line of GPUs will take place in May.

The postponement of the launch of the R600 cards is having various effects on the market. NVIDIA, for one, also chose to postpone the launch of its new cards of the G80 family leaving its partners with an unchanged product line and preventing them from officially launching any new hardware this year at CeBIT. Comparatively, though, manufacturers of NVIDIA-based parts are getting the better end of the deal as they can still market GeForce 8800 GTX products as the “only DirectX 10 hardware available.” Manufacturers who solely use Radeon parts, on the other hand, don’t have any new technology to market or introduce.

The AIB (Add In Board) partners of ATI that we met at CeBIT had very little to smile about. They weren’t able to release any new ATI technologies, and for the most part, as you will see in the article, they weren’t able to release any novelty products other than a couple of passive cooled parts and new cooler designs. In the past, ATI has been repeatedly criticized for its inability to honor its own roadmaps. You might remember the postponement of the Radeon X1800 cards which took place during the fall of 2005. The delay made it nearly impossible for many ATI partners to introduce new products that could compete with NVIDIA offerings that were already a generation ahead.

The same scenario is being repeated with R600 cards now. Fortunately, though, these manufacturers are only being directly impacted in the upper end performance segment. The fact that NVIDIA has chosen to delay its G8x until after CeBIT is a stroke of luck for ATI-AMD’s partners as they can still keep on releasing mainstream/value Radeon parts that are able to compete with G7x mainstream/value parts in terms of price and performance.

At CeBIT we were able to see working R600 cards along with the mainstream RV610 and entry level RV630. Although we are not allowed to show actually pictures of the parts, we still want to discuss some of the details we observed.

Starting with the R600, we were shown a retail version of the card, which is smaller than the 12-inch one that was shown at CES. Interestingly enough, the retail sample we were shown had only a single PCIe 6-pin power connector, a definite plus for a card that is targeted at the upper end market segment. For some time now there has been widespread speculation of whether the R600 will use a 65nm process in place of the originally planned 80nm. Although we are not entirely sure, and it has not been confirmed officially by AMD, we speculate that AMD was able to reduce the power consumption of the R600 by moving from 80nm to 65nm, thus meaning they only needed to put one PCIe 6-pin connector on the card. In addition, use of the smaller production might have also allowed AMD to increase the operating frequency of the R600. As the cards still haven’t been released though, the information might change in the future. The card we saw running had only a single PCI Express 6-pin power connector, but it might have actually been a mislabeled RV630 (Radeon X2600). Without official confirmation though, at the moment only time will be able to tell what the final specifications of the board will be.

The R600 card we were shown is completely covered by the cooling system which is comprised of a plastic case that covers the fan and the heatsink. Altogether, the heatsink is not extremely large, another indication that AMD might have opted on using a 65nm process in place of the 80nm process which we were expecting them to use. Again, however, this information hasn’t been confirmed by AMD.

For the RV630, the mainstream card, there is no additional power connector. The card also comes with two Crossfire connectors on the top of the PCB, meaning it supports native Crossfire.

The card that caught our attention the most was the RV610, AMD’s next generation value offering that supports DirectX 10. The card we were shown was quite small and it didn’t require additional power connectors. The card will only support software Crossfire, and not the clutter reducing native Crossfire which many of us have come to appreciate. The card will be relatively low cost, partly because it might be using a 65nm production process, and also because of the rather simplified architecture that entry level video cards use.

We believe the use of a 65nm process will give AMD an advantage over NVIDIA’s G80 GPUs, which use a 90nm process, by allowing it to reduce its manufacturing costs and also allowing it to offer aggressively priced DirectX 10 solutions. The various Add In Board partners we spoke to also believed the smaller production process would work in the favor of AMD and play a pivotal role for the company in the next generation card battle. They were also quite enthusiastic of the market opportunities for the new cards.

At the present moment, all technical and architectural characteristics, along with performance, availability, and pricing of the R600, RV610 and RV630 will be revealed in two months. In addition, in May we will finally be able to get a clear picture of the DirectX 10 cards of every market segment. In the meantime, we hope that software developers will begin releasing the first DirectX 10 based games to allow us to fully take advantage of DirectX 10 hardware.

The 8800 GTX and GTS have already been released, but NVIDIA’s entire 8800 series lineup has yet to be completed. In the next few weeks, the GeForce 8600 and GeForce 8300, whose technical characteristics have already been leaked on the Internet but not officially confirmed by NVIDIA, will be released. MSI was showing cards based on these GPUs, however, they did not specify their commercial names nor their technical characteristics to avoid violating the embargo date on these products.

The two cards here are expected to belong to the GeForce 8600 range. As is expected, the top of the board sports an SLI connector, while the boards also come integrated with DVI and HDMI connections for external displays. Although the GPU is covered by a sticker, it is still easy to see that the surface area is quite small meaning that NVIDIA has been able to cut production costs by getting a larger number of die per wafer.

The card with the red PCB is expected to be the GeForce 8500. The SLI connector, as is foreseeable, is present on the top left of the board and there is no 6-pin PCIe power connector. The cooling system is a single slot solution, and the board comes with VGA and DVI outputs.

To close the range of NVIDIA DirectX 10 offerings, MSI was displaying two cards which will belong to the GeForce 8300 family of products. In this case, there are no SLI connectors. Both boards, however, are also designed quite differently. The first one is meant to be a low profile card specifically designed for HTPCs, indicated by the HDMI connector and HDCP support. The other board is of standard size, and uses a dual-slot cooling system to allow higher operating frequencies than the NVIDIA reference line. A booth representative confirmed that the second card will come factory overclocked and that its name will end in "O.C. Edition.”

MSI was also showing a product that we have been looking forward to for the past couple of weeks since its original press release; the NX8800GTX OC Liquid. The card is based on the GeForce 8800 GTX GPU and it comes overclocked complete with a pre-installed liquid cooling system. Unfortunately we weren’t able to gain more details on the launch date of this product or its pricing, but be sure that we will have a review of it once it is released.

At CES in Las Vegas, which took place last January, Sapphire was showing a Radeon X1950 card which had two GPUs on one PCB. Two months later at CeBIT, Sapphire showed us two of these cards connected to each other through Crossfire, successfully creating a Quad-Crossfire system.

The two cards are quite large, in terms of both length and height. If they will ever actually be sold by Sapphire, they will definitely require a large case to actually hold them, unless of course you don’t use a case at all. The cards each have two Crossfire connectors, similar to what is used on the Radeon X1950Pro and Radeon X1650XT.

Each card needs two PCIe 6-pin connectors, therefore making them in a way similar to the NVIDIA GeForce 8800 GTX GPUs. Looking at the cards from the back, we can see that each separate GPU on the card is located at a different level to help accommodate the memory chips. The technical characteristics of the two GPUs on the boards are identical to those of the Radeon X1950PRO.

Sapphire was running 3D Mark 2006 on the Quad-Radeon system in loop. Performance, though, was not being displayed during the tests. They also failed to run any other 3D tests. The use of Quad Radeon X1950PROs probably has many conflicts with the current set of ATI drivers, so it is not very hard to imagine why Sapphire isn’t able to run many 3D applications on the Quad-GPU system. Quad-Crossfire technology will probably not make it to the market implemented on Radeon X1950PROs. Keeping in mind that the release of AMD’s DirectX 10 cards is right around the corner, it is much more likely that we will see the technology adopted for use with the new upper end DirectX 10 cards.

GeCube was showing a number of unique video cards this year at its exhibit. Despite the absence of a new GPU from ATI, GeCube was still able to come up with some interesting products to put on display that are different from what other manufacturers are offering.

Pictured above is a Radeon X1950XT card that uses an AGP 8x interface, something no other AIB partner of ATI is selling. The operating frequencies are identical to the PCI Express versions of the card, and it comes guaranteed to work at an overclocked GPU frequency of 675 MHz, up from 648 MHz. The cooling system, which is based on a TEC system, features a heatpipe technology heatsink and dual cooling fans.

The first time GeCube showed its own card with two ATI GPUs on the same PCB was at Cebit 2006. After one year the product has matured considerably throughout development, primarily thanks to the use of two Radeon X1650XT GPUs in place of the original X1600PRO GPUs. The unique aspect of this card is that when the two GPUs are connected to each other using Crossfire (still one board though), the card supports a total of 4 DVI connectors for external displays meaning that it is possible to have a quad-monitor setup using only a single video card.

The last novelty presented by GeCube was the above card based on the Radeon X1950XT GPU. The PCI Express 16x card, however, doesn’t use the standard Radeon X1950XT PCB, but instead uses the X1950PRO PCB. The use of the X1950PRO PCB allows the card to support Native Crossfire, which means a master/slave card along with the external Crossfire dongle is no longer needed.

Many partners of NVIDIA and ATI-AMD have introduced passive cooled cards for a long time. Passive cooling hasn’t only been applied to entry level cards. Thanks to more complex cooling solutions, more and more upper end performance segment cards are actually beginning to appear that feature completely silent cooling.

We took a look at three passively cooled Radeon cards from MSI. The first two are based on the X1650 GPU. Although their cooling solutions have been designed differently, their overall approach is to move heat towards the back of the PCB. The third card is based on the Radeon X1950PRO GPU. The card’s most prominent feature is the large heatsink which it comes installed with. The heatsink itself covers nearly the entire card. It is also interesting to notice that on the top right of the board MSI has placed a 4-pin molex connector for additional power in place of the newer 6-pin PCIe connector.

PowerColor and Club 3D have opted for the same type of passive cooling systems to use with their Radeon X1650 and Radeon X1950 cards. The cooling system has been designed by Artic Cooling, a company famous for its balance between efficiency and noise. In this case, four heatpipes are in contact with the GPU, which then lead to a large heatsink which covers the entire surface of the card.

ATI cards aren’t the only ones that were passively cooled at CeBIT. Gainward was using a thick heatsink for its Bliss 7600GT 256 SFX card, while Inno3D was also using an Artic Cooling solution, the iChill, for its GeForce 7950 GTs based GPUs.