by Alexey Stepin , Yaroslav Lyssenko
01/14/2009 | 04:05 PM
2008 was a tough year for Nvidia. The company had to suffer defeat after defeat, retreating in every sector of the discrete 3D graphics market before the sudden and energetic attack of AMD’s graphics department. Nvidia has got itself to blame for these misfortunes, however. The company had chosen a wrong development strategy for its GPUs, putting all effort into creating the G200. Nvidia had also delayed with the transition of its cores (including the previous-generation G92) to thinner manufacturing process.
The high potential of the G200 chip could not be untapped on 65nm tech process. Being a highly complex with its 1.4 billion transistors, the G200 just could not work at high frequencies. The frequency of the shader domain of the flagship GeForce GTX 280 was limited to 1.3GHz whereas the GeForce GTX 260 had to be clocked at 1242MHz only. For comparison, the shader domain of the original GeForce 8800 GTS (with a 90nm G80 core) used to be clocked at about the same frequency! And most unpleasantly, the new GPU was not always superior to the simpler RV770 chip from ATI.
That’s why the G200 needed to transition to newer and thinner tech process if it was to be competitive. This transition could increase the chip’s frequency potential while keeping its power consumption and heat dissipation within acceptable limits. And this could also pave the way for Nvidia’s counterpart of ATI’s Radeon HD 4870 X2. Creating a dual-GPU card with two 65nm G200 chips would result in an unacceptably hot and uneconomical product.
People at Nvidia realized all that better than anyone else. Losing one’s share of the discrete GPU market is easy but winning it back is a daunting task. As the result of Nvidia’s effort, the 55nm version of the G200 was created. It is also known under the codenames of G200b, GT200b, GT206 and some others. We will call it G200b. There is nothing new about the architecture of this chip. Like the G200, it incorporates 240 unified shader processors, 80 texture-mapping units and 32 raster back-ends. The only difference is that the G200b is manufactured on 55nm tech process. It is thus supposed to be cooler and more economical. Or, if its power consumption and heat dissipation are as high as those of the 65nm G200, the G200b should be faster.
The GeForce GTX 260 graphics card still uses cut-down chips with 216 ALUs, 72 TMUs and 28 raster back-ends. The G200 remains just as complex as before even in the 55nm version, so its manufacturing cost is quite high. Therefore, we guess that some cores installed on the GeForce GTX 260 Core 216 card either didn’t pass the frequency check or have defective subunits, which prevents Nvidia from using them for the GeForce GTX 285 and 295.
EVGA, one of Nvidia’s closest partners, was the first company to introduce graphics cards based on the new version of the G200 GPU. This privilege is a kind of a reward for loyalty: EVGA produces graphics cards based on Nvidia’s GPUs only. Other manufacturers are sure to follow the suit soon, too. But thanks to EVGA we’ve got the opportunity to check out the G200b right now using the EVGA GeForce GTX 260 Core 216 Superclocked card.
The product box seems to be an ordinary medium-size thing. However, the box is sealed into polyethylene film that looks like an indication of a high product status. The laconic design adds to this impression.
An orange strip against a black background is all the decoration here, but this minimalistic approach looks calm and respectable in comparison with all those robots, dragons and busty lasses in armor that you see on other product boxes. There is a widespread mistake on the packaging: the caption in the top left corner says “896MB DDR3,” but the card is equipped with GDDR3 memory which has little to do with DDR3. It is closer to DDR2 architecturally. The word Superclocked indicates factory overclocking and the sticker below it informs you that a copy of Far Cry 2 is included with the card.
There is a window in the back of the box through which you can see part of the PCB with the serial number sticker. Below the window there is a second sticker and text warning you that the serial numbers on both stickers must be identical. This is necessary for the lifetime warranty offered by EVGA as well as for the participation in the EVGA Step-Up program that allows replacing your card with a more advanced one (by paying the difference) in 90 days after the purchase. Well, we don’t guess that many people get tired of their new graphics cards so quickly. A graphics card usually serves for more than 90 days.
The packaging quality is traditionally high: EVGA uses a translucent plastic container instead of cardboard. The graphics card is fixed within the container, being perfectly protected against possible damage during transportation and storage. It may only be harmed if a storehouse loader runs over it, but no protection would help then. Besides the card, we found the following accessories in the box:
Even considering the rather high recommended price (over $250), the accessories are good, especially due to the copy of the popular shooter Far Cry 2. Our only gripe is about the lack of software for playing modern high-definition video formats, but the EVGA GeForce GTX 260 Core 216 Superclocked is positioned as a gaming solution. Moreover, such software is missing in most other graphics card kits.
Besides the drivers, the included CD contains an electronic version of the user manual and a couple of useful tools. One of them is the well-known Fraps (version 2.9.6) and another is EVGA Precision. The latter is an overclocking tool that seems to be based on the RivaTuner core. Besides controlling the graphics core and memory frequencies, this program allows you to adjust the speed of the cooler’s fan and keep track of the temperature of up to four GPUs in your system.
EVGA Precision allows you to create up to 10 profiles and assign hot buttons to them. It can output its data, including the frame rate, not only within the game (like Fraps) but also to the integrated display of the popular gaming keyboard Logitech G15. The program has a simple and functional interface, which makes it a handy everyday tool in an overclocker’s toolbox.
As for Far Cry 2, this game is the sequel to the legendary Far Cry that once set a new standard of graphics quality for first-person shooters. This game doesn’t need our recommendations as it should be familiar to every lover of this genre. We also use it as one of our tests. This is indeed a good gift for every gamer and EVGA has got our praises for that. Few graphics card vendors include new games with their products.
Thus, the packaging and accessories of the EVGA GeForce GTX 260 Core 216 Superclocked graphics card are top quality. The box looks attractive and contains everything necessary to use the card. The kit does not include obsolete cables and adapters but offers a new and popular game as a free bonus.
The card doesn’t seem to differ from the 65nm version of GeForce GTX 260/GTX 260 Core 216 but EVGA tried to make it original by putting a sticker on the cooler’s casing and painting the side of the casing red.
The card is 27 centimeters long and won’t suit compact system cases. This is a drawback because the opposing Radeon HD 4870 is only 23 centimeters long.
There are in fact more differences than you can see at first sight. We did not expect Nvidia to develop a new PCB design for the 55nm version of GeForce GTX 260 Core 216. It wouldn’t make sense because the card is in fact a cut-down version of GeForce GTX 280. However, Nvidia did develop it. We can see a lot of differences as soon as we remove the cooler.
First of all, the memory chips are now all located on the face side of the PCB and there are 14 of them. That is, this PCB design provides for a 448-bit memory bus without the opportunity of enlarging it to 512 bits. We don’t know the reason for Nvidia to develop the new PCB. Perhaps the company wanted to cut the manufacturing cost but we don’t think they win a lot with the new design. The updated PCB still looks very complex and expensive.
The power section has been revised considerably. The old version used a five-phase regulator that represented a cut-down variant of the seven-phase regulator of the GeForce GTX 280, but the 55nm GeForce GTX 260 has four phases in the GPU voltage regulator and there are different power transistors in them. The bad news for the owners of liquid cooling systems or nonstandard air coolers: monolithic cooling solutions developed for the 65nm GeForce GTX 200 series and designed to cool not only the GPU and memory chips but also the power circuit elements will not fit the 55nm cards because their power elements are located differently.
The voltage regulator is based on a four-phase PWM controller NCP5388 located on the reverse side of the PCB. Next to it, there is a tiny mysterious chip marked as BR=AL U07. It must be responsible for controlling the dedicated two-phase memory voltage regulator. The card has two external power connectors. Both are of the 6-pin PCIe 1.0 variety and have a load capacity of 75W. The metallic frame around the GPU is one more difference whereas the left part of the PCB, with the interface connectors and the NVIO chip, has been left intact.
The EVGA card uses GDDR3 chips from Samsung. Marked as K4J52324QH-HJ1A, the chips have a rated frequency of 1000 (2000) MHz and a voltage of 1.9V.
There are 14 of them here, offering an 896MB memory bank with a 448-bit memory bus. According to the official specifications from Nvidia, the GeForce GTX 260 has a memory frequency of 1000 (2000) MHz irrespective of the GPU version. This provides a memory bandwidth of 112GBps. But EVGA pre-overclocked the memory chips to 1053 (2106) MHz, increasing the bandwidth to 117.9GBps. This is somewhat higher than the Radeon HD 4870’s memory bandwidth of 115.2GBps but leaves no room for further overclocking considering the rated frequency of the chips. We’ll check out the card’s overclockability in the appropriate section of the review, though.
We could not check visually how smaller the G200b was. Like the old version of the chip, the new version is equipped with a metallic heat-spreading cap. We did not dare remove that cap as we needed the card for our tests. Anyway, the numbers speak for themselves: the new G200 is as large as 470 sq. mm as opposed to the old version’s 576 sq. mm (the RV770 is far more compact, measuring 260 sq. mm only, but also much simpler in terms of transistors). The GPU is marked as G200-103-B2, so you can easily tell it from the old version that used to be marked as G200-100-A2.
According to the official specs, the GeForce GTX 260 Core 216 has 576MHz and 1242MHz frequencies for the main and shader domains, respectively, but the GPU of the EVGA card is pre-overclocked to 625MHz and 1350MHz. That’s not much, and we can’t say if the frequency potential of the G200b is higher than that of the G200. We have seen 65nm versions of GeForce GTX 260 Core 216 pre-overclocked to higher frequencies, so we will try to overclock the EVGA card further shortly.
The GPU configuration is standard for GeForce GTX 260 Core 216: 216 shader processors, 72 texture processors, and 28 raster back-ends. If the card has good overclockability, it can deliver the performance of the GeForce GTX 280 and more.
The card is equipped with two dual-link DVI-I ports (with support for resolutions up to 2560x1600 inclusive), a universal analog video output, two SLI connectors (for building a graphics subsystem out of two or three such cards), and a two-pin S/PDIF connector (to translate an external S/PDIF audio stream into HDMI using the included cable).
The cooling system of the 55nm GeForce GTX 260 Core 216 has been updated due to the new PCB design and the more economical graphics core.
The most notable difference is the reduced size of the heatsink. It has become shorter because it now lacks the section that used to be located in front of the mounting bracket. The heat pipe that used to transfer the heat from the power circuit elements is now missing whereas the cooler’s base that contacts with the GPU die has become much smaller. The pipe that cools the NVIO chip is present, though.
These revisions must have been meant to reduce the manufacturing cost of the cooler. We don’t see any other reason to cut down the configuration of the time-tested cooler design. Of course, the revised cooler is going to have lower performance, but the G200b should produce less heat. We will check this out in the next section.
Otherwise, the concept has remained the same: the copper heat-exchanger takes heat off the GPU and transfers it to the heatsink by means of heat pipes. Traditional dark-gray thermal grease is used as a thermal interface. The heatsink is cooled by a blower, and the hot air is exhausted out of the system case through the slits in the card’s mounting bracket. The other elements that require cooling such as memory chips, NVIO chip, and power transistors of the power circuit contact with the juts in the cooler’s aluminum base through fabric pads soaked in white thermal grease.
The cooler looks good, even though simpler than the original version developed for the GeForce GTX 200 series. Its performance should be lower, but the difference will hardly be noticeable due to the use of the 55nm version of the G200 core. We’ll check this out right now.
We measured the power consumption of the 55nm GeForce GTX 260 Core 216 graphics card on the following testbed to see if it is more economical than the older 65nm version:
The 3D load was created by means of the first SM3.0/HDR test from 3DMark06 running in a loop at 1600x1200 with 4x FSAA and 16x AF. The Peak 2D mode was emulated by means of the 2D Transparent Windows test from PCMark05. This test is important as it simulates the user’s working with application windows whereas Windows Vista’s Aero interface uses 3D features.
Click to enlarge
The G200 chip and the GeForce GTX 260 card have become more economical thanks to the 55nm tech process. The difference is small in idle mode, but the card’s power draw in 3D mode is only 34 watts. Thus, G200b-based solutions have surpassed the Radeon HD 4870 1GB in this respect. ATI should think about ways to improve the power efficiency of its products, especially dual-core ones.
The load is distributed unequally between the two external power connectors of the new version of the GeForce GTX 260 Core 216 as opposed to the older version that uses the PCB design and power circuit developed for the GeForce GTX 280. The connector that is nearer to the end of the card is loaded by 10-13 watts more. Anyway, this is no reason to worry because the load is far lower than the permissible 75W.
The 55nm version of the GeForce GTX 260 Core 216 has the same frequencies as the 65nm version: 576MHz for the main GPU domain and 1240MHz for the shader domain. The EVGA GeForce GTX 260 Core 216 Superclocked card is pre-overclocked by the manufacturer to 625/1350MHz but we decided to overclock the card more to check out the frequency potential of the G200 chip after its transition to 55nm tech process. Without any special means (such as volt-modding or replacement of the cooler), we achieved frequencies of 715/1541MHz for the graphics core and 1150 (2300) MHz for the memory chips. That’s not bad for a 1.4-billion-transistor chip, especially as our 65nm version of the GeForce GTX 260 Core 216 could only overclocked to 650/1400MHz. The frequency gain is 10%, which may make the card competitive to the GeForce GTX 280 that has 240 ALUs, 80 TMUs and 32 RBEs against 216, 72 and 28 such subunits of the GeForce GTX 260 Core 216.
We monitored the graphics card’s frequency with RivaTuner:
The 55nm and 65nm versions have the same GPU temperature at the same frequencies. This must be due to the simplified cooler of the former card. Anyway, when we overclocked the EVGA card more, the top GPU temperature was not higher than that of the GeForce GTX 280. Note that every GeForce GTX 200 card automatically lowers the GPU clock rates to 300/600MHz and the memory frequency to 100 (200) MHz in 2D mode to keep the temperature and noise lower.
Although the EVGA card’s cooler is somewhat different, it produces the same amount of noise as the reference GeForce GTX 280 cooler. This is also due to the rather high level of noise from our testbed at large (43dBA with a passively cooled graphics card inside). Anyway, Nvidia’s cooler is still one of the best in the industry, combining high cooling performance with low noise. It is not exactly silent, but the hissing of its airflow is not irritating at all.
We are going to investigate the performance of our EVGA GeForce GTX 260 Core 216 Superclocked graphics card using the following testbed:
The graphics card drivers are configured in the same way as before: to provide the highest possible quality of texture filtering and to minimize the effect of software optimizations used by default by both: AMD/ATI and Nvidia. Also, to ensure maximum image quality, we enabled transparent texture filtering, and we used multisampling mode for both graphics architectures, because ATI solutions do not support supersampling for this function. As a result, our ATI and Nvidia driver settings looked as follows:
We made a lot of changes to the list of games and benchmarks we normally use for our tests in order to meet the contemporary standards. As a result, it currently includes the following titles:
First-Person 3D Shooters
Third-Person 3D Shooters
We selected the highest possible level of detail in each game using standard tools provided by the game itself from the gaming menu. The games configuration files weren’t modified in any way, because the ordinary user doesn’t have to know how to do it. We made a few exceptions for selected games if that was necessary. We are going to specifically dwell on each exception like that later on in our article.
Besides EVGA GeForce GTX 260 Core 216 Superclocked we have also included the following graphics accelerators to participate in our test session:
We ran our tests in the following resolutions: 1280x1024, 1680x1050, 1920x1200 and 2560x1600. Everywhere, where it was possible we ran the tests with enabled MSAA 4x antialiasing and anisotropic filtering 16x. We enabled them from the game’s menu. If this was not possible, we forced them using the appropriate driver settings of ATI Catalyst and Nvidia ForceWare drivers. As we have already said, we didn’t modify the games configurations files.
Performance was measured with the games’ own tools and the original demos were recorded if possible. Otherwise, the performance was measured manually with Fraps utility version 2.9.6. In the latter case we ran the test three times and took the average of the three for the performance charts. We measured not only the average speed, but also the minimum speed of the cards where possible.
EVGA’s factory overclocking is not enough to make the card competitive to the GeForce GTX 280, but our additional overclocking achieves that even at the resolution of 2560x1600, notwithstanding the narrower memory bus. Moreover, the bottom speed is high enough for comfortable play at the highest resolution. The overclocked card from EVGA is no worse than the Radeon HD 4850 X2 but consumes far less power.
We can see the same picture in Crysis Warhead. Our overclocking makes the EVGA card as fast as the GeForce GTX 280, except for the resolution of 2560x1600 where the narrower memory bus and the smaller mount of graphics memory (896MB against 1024MB) of the EVGA card affect its performance.
The frame rate is fixed at 30fps in this game as this is the rate at which all events are synchronized during networked play. We disabled this limit in the game console for the sake of comparing the cards. The game’s built-in benchmarking options do not provide information about the bottom speed, so there is no such info in the diagrams.
There is hardly any difference between the GeForce GTX 200 series cards in this test. When overclocked, the EVGA card is as fast as the GeForce GTX 280 and even ahead of the latter at 2560x1600. The frame rate is always higher than necessary for playing this game.
The original EVGA GeForce GTX 260 Core 216 Superclocked is somewhat slower than the GeForce GTX 280, but it closes the gap when overclocked to 715/1641MHz GPU and 1150MHz memory (there is still a small gap in terms of bottom speed, though).
To achieve a playable speed in this game we disabled FSAA and such resource-consuming options as Sun rays, Wet surfaces and Volumetric Smoke. We use the Enhanced full dynamic lighting (DX10) mode for our test and additionally enabled the DirectX 10.1 mode for the ATI cards.
The overclocked card from EVGA behaves in the same way in Clear Sky, too. It is as fast as the GeForce GTX 280 in terms of average frame rate thanks to the increased frequencies. But its bottom speed is lower because it has fewer functional subunits and a narrower memory bus.
Dead Space confirms our point that the higher clock rates cannot make up for the lack of ALUs, TMUs and raster back-ends: the overclocked EVGA is 2-5% ahead of the GeForce GTX 280 in average frame rate but its bottom speed is no higher than at the default frequencies. Moreover, overclocking is not enough at 2560x1600 where the GeForce GTX 280 goes ahead again. The game has modest system requirements while the cards are very fast. The mentioned differences cannot be spotted with a naked eye.
The EVGA card is faster than the GeForce GTX 280 in terms of average speed but has the same bottom speed. The only exception is the resolution of 2560x1600 where the EVGA delivers a higher bottom speed than the flagship of the GeForce series.
The game does not allow to set Texture Quality at higher than Medium for graphics cards with 512MB of graphics memory whereas the maximum safe value of View Distance is 32. So, we had to select the mentioned values for our test. The other options were set at their maximums. The game’s built-in benchmarking options do not provide information about the bottom speed, so there is no such info in the diagrams.
It is only at 1920x1200 and 2560x1600 that we can see any difference between the tested graphics cards. The overclocked EVGA delivers the same average frame rate as the GeForce GTX 280 at 1920x1200 and even overtakes the latter at 2560x1600. It also beats the Radeon HD 4850 X2. Judging by the results of the Radeon HD 4870 1GB, the game prefers Nvidia’s GPUs that have a lot of texture processors.
Starting from our next review we will use other settings in this game, oriented at graphics cards with 896MB or more memory.
ATI’s solutions are beyond competition, but Nvidia’s G200-based card feel at ease at every resolution including 2560x1600, too. Interestingly, the EVGA GeForce GTX 260 Core 216 Superclocked is no worse than the GeForce GTX 280 even at the default frequencies. When overclocked, it is 8-10% ahead of the latter.
Like in many previous tests, the overclocked EVGA card is competitive to the GeForce GTX 280 in average and bottom speed. The Radeon HD 4870 1GB is slower than the G200/G200b-based solutions in this test. They are only overtaken by the hotter and noisier Radeon HD 4850 X2.
The overclocked card from EVGA is second only to ATI’s dual-core monster at 2560x1600. Otherwise, we’ve got the same picture as in the previous tests: the factory overclocking is not enough to compete with the GeForce GTX 280. The card has to be overclocked more to deliver the same average frame rate, but its bottom speed doesn’t grow up much then.
The game obviously prefers ATI’s solutions: just take a look at the bottom speeds at resolutions higher than 1280x1024. Overclocking doesn’t help the EVGA card much. Its bottom speed grows up to 22fps only at 1680x1050, which is not enough to ensure smooth gameplay. Interestingly, the modest factory overclocking by EVGA helps the GeForce GTX 260 Core 216 Superclocked outperform the GeForce GTX 280 at 1920x1200 and 2560x1600.
The game has a built-in frame rate limiter set at 30fps.
Nvidia’s card are unacceptably slow in Red Alert 3 if you enable 4x FSAA. Hopefully, this problem will be corrected with the next driver update. Anyway, the overclocked card from EVGA behaves just like in the previous tests.
The game has an integrated frame rate limiter set at 30fps. It doesn’t support FSAA.
This test does not give us useful information. All the cards hit the frame rate limit. We will replace this test with a more informative one in our future reviews.
Our overclocking of the EVGA card produces an impressive result: this card beats the Radeon HD 4850 X2 at 1680x1050. Besides ATI’s dual-core solution, it is the only card to provide a comfortable speed at 1920x1200. The largest gap from the GeForce GTX 280 amounts to 14%, which is a superb result considering the difference in the number of subunits as well as in the performance and amount of graphics memory.
The EVGA is not exceptional in any of the 3DMark06 tests irrespective of its frequencies. Perhaps it is limited by the hardware configuration of the GeForce GTX 260 Core 216.
We minimize the CPU’s influence by using the Extreme profile (1920x1200, 4x FSAA and anisotropic filtering). We also publish the results of the individual tests across all display resolutions to provide a full picture.
3DMark Vantage draws a completely different picture: the overclocked EVGA outperforms the GeForce GTX 280 and takes first place among all the participating graphics cards!
It is at the resolution of 1280x1024 that our overclocking is the most rewarding. The gap between the EVGA card and the GeForce GTX 280 shrinks as the resolution is increased. After all, the latter has a faster memory subsystem, and the higher GPU frequencies of the EVGA can hardly outweigh this advantage.
It’s different in the second test: the EVGA is somewhat slower than the GeForce GTX 280 at the default frequencies and equals it at the overclocked frequencies. The lack of texture processors seems to be compensated by the higher frequency of their operation. Note that both solutions from Nvidia are far slower than the Radeon HD 4850 X2 in this test.
Our tests have revealed no difference in performance between the 55nm and 65nm versions of GeForce GTX 260 Core 216 at the reference GPU and memory frequencies. However, the factory overclocking of the EVGA card ensured it an advantage of 4.3%-6.4% depending on the resolution. And our overclocking added 8-9% more to the card’s speed.
As a result, the EVGA GeForce GTX 260 Core 216 Superclocked showed itself a worthy opponent to the ATI Radeon HD 4870 1GB. The EVGA card is faster in many tests, but the average advantage is only 5% (at resolutions no higher than 1680x1050) due to its loss in such games as Fallout 3, Race Driver: GRID, X³: Terran Conflict and Red Alert 3. Thus, the choice between the two cards should be based on the gamer’s personal preferences.
The EVGA card does not equal the GeForce GTX 280 at its default frequencies, but it showed good overclockability thanks to the new 55nm version of the G200 core. When overclocked, it overtook and even outperformed the flagship of the GTX 200 series in nearly every test.
The average advantage of the EVGA GeForce GTX 260 Core 216 overclocked to 715/1541/2300MHz varies from 3.1 to 3.8%, and the biggest effect could be observed in Devil May Cry 4 (1280x1024, 10.7% advantage), World in Conflict (1680x1050, 10.8% advantage) and X³: Terran Conflict (1920х1200 and 2560x1600, 16.2 % and 19.3 %, respectively). This is very good considering the much lower price in comparison with the GeForce GTX 280. Thus, purchasing the latter card makes little sense after the arrival of the 55nm GeForce GTX 260 Core 216, but the upcoming GeForce GTX 285, another product based on the 55nm G200, is going to have higher frequency potential, too.
Talking about the specific product, the EVGA GeForce GTX 260 Core 216 Superclocked delivers high performance and boasts good overclockability. It also comes with good accessories that include a copy of Far Cry 2, handy overclocking tool, lifetime warranty, and the opportunity to participate in the EVGA Step-Up program. The only downside is the high price, typical of EVGA solutions, especially as the official price of the Radeon HD 4870 with 512MB of GDDR5 memory has already dropped below $200 while the 1024MB version of the card costs only $239.