Just a few weeks ago ATI released their first video card that used a GPU constructed under an 80nm production process. We are referring to the X1950 PRO, ATI’s newest midrange video card. The X1950 PRO, which uses an RV570 GPU, introduced a number of new innovations. Most interesting was the new method of handling Crossfire technology. ATI’s new approach to Crossfire which debuted with the release of the X1950 PRO has done away with the requirements of master cards. In addition, ATI has opted upon getting rid of the external Crossfire dongle and instead chosen a much more NVIDIA-like method, the use internal bridges to connect two cards in Crossfire mode. Yes, we said “bridges”. Contrary to SLI, the new Crossfire approach uses not one, but two bridges to connect multiple GPUs. The reason behind this is that multiple bridges will supposedly offer more advanced bandwidth than would be possible using just one.

Today our article is about the X1650 XT, ATI’s newest video card part of the X1650 family. The X1650 XT is based on the RV560 core, which is ATI’s second GPU to be manufactured under an 80nm manufacturing process. The ATI Radeon X1650 XT also uses the same revised Crossfire approach as the X1950 PRO.

Radeon X1650XT in Crossfire configuration

The Radeon X1650 XT is aimed towards the lower mid-range sector of the market. We can better understand the market segment this card fits into by looking at its price. The MSRP for the X1650 XT is 149 USD. By looking at competing NVIDIA solutions, the one most similar in price is the 7600 GT, which is available for 10 USD less at a price of $139. Because of their similar prices, the X1650 XT’s main NVIDIA competitor is naturally the 7600 GT. Both are aimed at the same market segment and are similar in terms of features. The card the X1650 XT is replacing, the X1650 PRO, is now available at the budget price of 99 USD. The GeForce 7900 GS, another potential competitor, has been cut down to the priced of 109 USD.

As can be seen in the table, all of the video cards have a 128bit bus width. The average size of the video memory is 256 Mbytes, however, some solutions from partners of NVIDIA or ATI might have the cards come with 512 Mbytes of video memory.

The main difference between the X1650 XT and the X1650 PRO is the difference in GPUs. For the X1650 XT, ATI has doubled the pixel shading units, going from 12 to 24, and also doubled the amount of TMU and ROPs. Memory bandwidth, however, has not been changed. Surprising enough, the core and memory clock speeds for the X1650 XT have also been reduced. The core clock has been reduced 10 MHz and the memory clock has been reduced 20 MHz. Despite this, for the most part the X1650 XT has a very large number of technical improvements over the X1650 PRO. From taking a look at the X1650 XT’s specifications, we can already tell that it will be able to easily outperform its predecessor the X1650 PRO.

Top: GeForce 7600GT; Bottom: Radeon X1650XT

The reference card of the X1650 XT that ATI provided is smaller when compared to the 7600 GT. As can be expected, the X1650 XT is based on a red PCB. Since the card is not a heavy duty high-end video card and can receive all of its power from the PCI-E 16x slot, it does not require an addition PCI-E 6 pin power connector. As can be seen in the photo, the X1650 XT has a large cooling system that covers much of the video card.

In the above photograph that shows a posterior view of the X1650 XT, the two Crossfire bridge connectors can be seen. The connectors are identical to that of the Radeon X1950 PRO. In our previous article over the X1950 PRO we learned from testing that it is required to use both bridges in order for Crossfire to work. It is also required to use identical models, so for instance we cannot have a successful Crossfire implementation by linking together the Radeon X1950 PRO and the Radeon X1650 XT.

After removing the cooling solution and heatsink, we took a look at the RV560 core. Using TSMC’s 80nm technology has allowed ATI to reduce heat in addition to cutting production costs of the GPU. Constructing under an 80nm process also allows for reductions in the dimensions of the GPU’s surface. The size of the GPU’s surface is 238 square millimeters. These dimensions are identical to that of the RV570 core which was reviewed a little while back. The RV570 core is also constructed under an 80nm process. Keeping this in mind, it is possible that the RV560 and RV570 are in reality based on the same base architecture; ATI might have just simply disabled some functions for the RV560. The fact that the RV570 has exactly 50% more shader units, ROPs, and TMUs than the RV560 only serves as more proof for this possibility.

We are concentrating the analysis of the X1650 XT against its main NVIDIA competitor, the 7600GT. Both cards share nearly the same MSRP, the X1650 XT being at $150 and the 7600 GT at $139. Naturally, both cards are also aimed towards the same market sector.

The test system we used was equipped with a Socket 939 AMD Athlon 64 X2 4800+. The system also had 2 Gbytes of DDR400 memory. We tested with resolutions of 1024x768, 1280x1024, 1600x1200 with the following filters enabled:

• 2X AA/ 8X Anisotropic
• 4x AA/ 16x Anisotropic

We didn’t conduct any tests without filters simply because we believe that our readers are more interested in the potential of the video card. We believe you would rather want to know how the video card performs with better image quality with use of at least 2x anti aliasing and 8x anisotropic filtering.

Benchmark Lineup

• 3DMark 2006 (synthetic test)
• Far Cry
• Splinter Cell: Chaos Theory
• Doom 3
• Serious Sam 2
• Call Of Duty 2
• Fear
• Prey

From the performance seen in the above graph, it is evident that the GeForce 7600GT performs better in the fill rate tests.

In the above benchmark, the 7600GT performs better in simple vertex shading tests. Moving on to complex vertex shading, however, the X1650 XT gains the advantage.

In the pixel shader tests, both cards perform very similar to each other. In the next test, Perlin Noise 3.0, the Radeon X1650 XT clearly performs better compared to the 7600 GT.

Far Cry performs better with the 7600 GT in all three test resolutions when Anisotropic filtering is set at 8x and FSAA at 2x. Even though the advantage is slight, when FSAA is set at 4x and anisotropic filtering at 16x, the picture slightly changes to favor the Radeon X1650 XT at the higher resolutions of 1600x1200 and 1280x1024. Looking at the picture as a whole, both cards perform pretty much equal to each other.

By now, it is well known that Doom 3 favors NVIDIA hardware over ATI. This is due to the fact that games based on the OpenGL API usually work better with NVIDIA architectures. The performance difference is between the NVIDIA solution over the ATI one is evident the above test. With filters set at 4x/8x (AA/AF) and High Quality enabled, Doom 3 has better performance with the 7600 GT. In all 3 test resolutions the 7600 GT always holds around a 10% advantage over the X1650 XT.

As we have been seeing seen with the game benchmarks run so far, the performance difference between the Radeon X1650 XT and GeForce 7600 GT is not a very large one. In no tests run so far have we seen either of the cards take a clear advantage in terms of performance. This is a general trend that repeats quite often throughout the remainder of this article.

The performance displayed by the cards stays true to the general trend. There is a slender performance difference between the cards and the X1650 XT has a light margin of advantage, the highest of which grazes 10% when running at 1600x1200 with AA 4x and FSAA 16x. Despite this, the advantage is too low to declare a true winner in for Serious Sam 2. With HDR enabled, the difference in FPS widens between the two cards and the X1650 XT begins to perform a bit stronger than the 7600 GT than the first Serious Sam 2 test. From the results above, we have to say that neither card is playable at 1600x1200. The frame rate is far too low, in our opinion, for someone to properly enjoy a first person shooter game.

Prey mosta un sostanziale allineamento tra le due schede video, con un lieve margine di vantaggio per la soluzione NVIDIA ma complessivamente trascurabile.

Yet again, both cards perform very similar to each other in the above tests. At times the performance differences can literally be less than 1 FPS. Thus it is only natural to say that in neither of the above tests we can declare a true performance winner between the two cards. It is also essential to note that neither of the two video cards offers playable performance at higher resolutions with anti-aliasing at 4x and anisotropic filtering at 16x.

Power consumption is an essential element of any review because it greatly contributes to the final judgment of a video card. Often this aspect of performance is overlooked; however, it makes a large impact on customer satisfaction.

Consumption of both cards was measured with a power monitor. We Analyzed consumption of the whole system under identical conditions, the only difference being the cards themselves. The “Idle” test was carried out using a blank desktop in Microsoft Windows XP while the test under full load was carried out with F.E.A.R. running at AA 4x and Anisotropic filtering at 16x. The GeForce 7600 GT consumed less power under idle and full load. The RV560 core, as the above tests reveal, had very similar power consumption compared to the RV570 core giving the theory the RV560 is only a RV570 with some calculation units disabled further backing.  

Overclocking tests were conducted using the integrated Catalyst utility. We would have used ATI Tool to overclock the video card, but we weren’t able to since the utility does not yet have support for the RV560 GPU.

The most stable clock we were able to reach in our testing was 612 MHz GPU and 770 MHz Memory.

Overclocked, we ran some more gaming tests with the X1650 XT using Prey. The results it gave us was interesting. With faster clock speeds, the X1650 XT was able to have more FPS than it would have with stock settings. In addition, the overclocked X1650 XT also had slightly more FPS than the 7600 GT in all tests.

Conclusion

The match-up between the 7600 GT and X1650 XT is a close one. The performance both cards offer is very similar to each other. Keeping this in mind, it is difficult to actually declare a winner between the two cards. The performance offered by the X1650 XT is pretty good considering it is aimed towards the lower-midrange sector of the market. The X1650 XT was flexible in gaming and was able to run most of the games, with filters on, at 1280x1024 at smooth frame rates. Although this card wasn’t able to run a couple of games at 1600x1200 with smooth frame rates, it is understandable in our opinion keeping in mind the card is aimed towards a market sector where most people aren’t hardcore gaming enthusiasts who always play with the highest resolutions available to them.

On ATI’s end, it seems that they are still having some trouble actually launching their products. The launch of the X1650 XT so far has only been one on paper. Quite possibly because ATI is still in the process of merging with AMD and things might be a bit hectic over there, ATI just hasn’t been able to provide manufacturers with enough parts yet. From what we are being told, X1650 XT cards should start actually appearing for sale mid-November.

The MSRP that ATI has set for the Radeon X1650 XT is $149. If the video card actually sells for this price, it would be a good buy. Unfortunately, that might not be the case. What we have seen with the X1950 PRO so far because of its limited availability to manufacturers prices have shot up. Currently, NewEgg is selling the card for $274.99, around $75 more than its MSRP. We hope that that will not be the case for the X1650 XT, but at this moment it is hard to tell. The projected price tag of $149 makes the X1650 XT a decent buy, but if it hits the $200 mark there wouldn’t be much point in buying the card when the 7600 GT can be easily found for less than $150.

We hope that this paper launch from ATI will only turn up to be a minor mess-up for the Canada-based company. With AMD now controlling the direction of both companies, we hope to see more steps being taken in order to ensure product availability at launch. We have high hopes that ATI will rectify this situation soon and that in their upcoming launches we will see actual hardware available on the market.