by Alexey Stepin , Yaroslav Lyssenko
06/30/2009 | 01:46 PM
As you know from our reviews, Nvidia had responded to the release of the ATI Radeon HD 4890 by announcing its GeForce GTX 275. This was a lucky move. Being a cut-down version of GeForce GTX 285, the new graphics card showed itself a worthy rival to the opposing Radeon and beat the latter in a number of games. However, the GeForce GTX 275 was inferior to the Radeon HD 4890 in one parameter.
We mean the amount of local graphics memory. ATI’s card features 1024 megabytes of ultra-fast GDDR5 memory accessed across a 256-bit bus whereas Nvidia economized on the RBE subsystem of its GeForce GTX 275 and cut down its memory subsystem as well. As a result, the GeForce GTX 275 has seven rather than eight 64-bit memory controllers and, consequently, a 448-bit memory bus and 896MB of graphics memory (with standard 512Mb GDDR3 chips). We can’t say that this affects the new card’s performance greatly because most games are more than satisfied with 896MB. And where this amount of graphics memory proves to be not enough, the card’s GPU performance is usually below playable.
Some time ago we posted a review of a GeForce GTS 250 equipped with as much as 2 gigabytes of graphics memory. However, the Gainward GTS250 2048MB did not have any advantage over Nvidia’s reference card with 1GB of memory. In that review we also supposed that large amounts of graphics memory would not be demanded in near future. There are more and more multi-platform game projects whose developers take into account the rather modest hardware specs of today’s gaming consoles and introduce appropriate optimizations. And the next generation of consoles from Microsoft, Sony and Nintendo is not going to be released soon.
However, the formal inferiority of the GeForce GTX 275 to the Radeon HD 4890 was obvious and inexperienced users could be misguided by the numbers: “1024MB” looks far more appealing on the product box than “896MB”.
This problem could be easily solved by replacing 512Mb with 1GB chips, though. As a result, the GeForce GTX 275 would have more memory than the Radeon HD 4890 and even than the dual-processor flagships from both ATI and Nvidia. Versions of GeForce GTX 275 with a double amount of memory (1792MB instead of 896MB) were sure to come out. One such product has been released by EVGA whose products often come to our labs and get positive reviews. Let’s see if the EVGA GeForce GTX 275 1792MB can carry this tradition on.
The graphics card comes in EVGA’s unified box that has already earned our praises for its restrained and stern design. As usual, it is the color of the band in the top part of the box that differentiates this card from other models. This time it is yellow and has a pattern made out of EVGA logotypes.
Otherwise, the box is no different from other EVGA products. The familiar mistake regarding the memory type is still here, too. The card is actually equipped with 1792 megabytes of GDDR3 rather than DDR3 memory. As usual, there is a window in the back of the box for you to compare the serial number stickers on the box and graphics card’s PCB in order to make sure that the product is authentic and eligible for the EVGA 90 Day Step-Up program.
The card is fixed in a blister pack and its accessories lie in the depressions of the pack. The device is thus protected against troubles during transportation and storage. It comes with the following accessories:
Besides, the box contains a 20% discount coupon for a copy of Mirror’s Edge and an advertisement of the EVGA Killer Xeno network adapter. The latter is quite an interesting product that is going to be covered in our upcoming report.
Besides drivers, the disc contains a couple of handy tools: Fraps and EVGA Precision. Unfortunately, the card’s power circuit is based on a controller from On Semiconductor rather than from Volterra and is not supported by EVGA’s Voltage Tuner, a tool for software control over the GPU voltage.
The packaging and accessories of the EVGA GeForce GTX 275 1792MB are good overall although a DVI-I → HDMI adapter is missing. The discount coupon is good, but a copy of Mirror’s Edge would be even better.
The EVGA GeForce GTX 275 1792MB is a copy of Nvidia’s reference card with twice the amount of memory chips. The PCB revisions even coincide: both copies of GeForce GTX 275 are marked as “180-10897-****-B01”.
The power circuit controller is based on an ADP4100 chip from ON Semiconductor. It controls a six-phase voltage regulator with three MOSFETs in each phase. The memory voltage regulator is based on uPI Semiconductor UP6161N and ENE Technology P2349WF chips (the latter is similar to the Intersil 6549CBZ). External power is connected via two 6-pin PCIe 1.0 plugs. The PCB does not provide for the installation of an 8-pin PCIe 2.0 connector.
Like in the standard version of GeForce GTX 275, the PCB carries 14 memory chips with a 448-bit interface. Each chip is 1024 rather than 512Mb, so the total amount of graphics memory is 1792 rather than 896MB. Despite this difference, the memory works at the reference frequency of 1134 (2268) MHz, ensuring a peak bandwidth of 127GBps.
The GPU is not pre-overclocked, either. Its main domain is clocked at 633MHz and its shader domain, at 1404MHz. The RBE and memory controller subsystem is cut down in comparison with the GeForce GTX 285. This card has 28 RBEs while the GTX 285 has 32. There is no difference between them otherwise. Both have 80 texture and 240 unified shader processors.
As opposed to an earlier-tested GeForce GTX 275 that lacked analog video output, the EVGA GeForce GTX 275 1792MB offers a full selection of interfaces including the above-mentioned universal 7-pin mini-DIN port, two DVI-I connectors, two MIO connectors, and a 2-pin S/PDIF.
The graphics card from EVGA is equipped with Nvidia’s cooler, so we won’t discuss its design in detail. We can just remind you briefly that the cooler consists of a composite aluminum heatsink and a copper heat-exchanger that are connected with heat pipes. The cooler is covered with a plastic casing and has a blower. The hot air is exhausted out of the system case through the slits in the graphics card’s mounting bracket. The whole arrangement is secured on the PCB with numerous spring-loaded screws – there is no risk of any misalignment. This is a time-tested design that can ensure acceptable cooling performance at a comfortable level of noise.
Although the EVGA only differs from the reference GeForce GTX 275 with the double amount of graphics memory, we measured its power consumption to check out the effect of the memory chips on the card’s power draw. The test was performed on a special testbed configured like follows:
Following our standard procedure, the 3D load was created by the first SM3.0/HDR test from 3DMark06 running in a loop at 1600x1200 with forced 4x FSAA and 16x AF. The 2D load was emulated by the 2D Transparent Windows test from PCMark05. The graphics card worked at its default GPU and memory frequencies.
Click to enlarge
Predictably, the power consumption is not much different. The results are good, yet the EVGA card, like every other version of GeForce GTX 275, is still far inferior to the Radeon HD 4890 in terms of power efficiency.
The temperature of the EVGA GeForce GTX 275 is similar to that of the reference card, too.
The GPU temperature is higher by a mere 2°C, which must be due to the summer weather and higher room temperature in our labs. As we tested the graphics cards under real-life conditions, i.e. under the same conditions that they are going to be used by gamers, the numbers above should be considered as just approximate.
Using the same cooler, the EVGA GeForce GTX 275 was no different from Nvidia’s reference card in terms of noisiness, either.
Our overclocking attempt was quite a success:
Without any special overclocking methods we increased the GPU frequency to 720/1597MHz and the memory chips to 1184 (2368) MHz. This addition to the default frequencies is big enough to affect the card’s performance in games, so we will test it at the overclocked frequencies, too. We even suspect that overclocking will have a bigger effect on the results of the EVGA GeForce GTX 275 1792MB than its 1792 megabytes of graphics memory. Let’s check this out right now!
We are going to investigate the performance of EVGA GeForce GTX 275 1792MB graphics card using the following testbed:
The graphics card drivers were configured in the following way:
ATI Catalyst:
Nvidia GeForce:
The list of benchmarks includes the following gaming titles and synthetic tests:
First-Person 3D Shooters
Third-Person 3D Shooters
RPG
Simulators
Strategies
Semi-synthetic Benchmarks
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 275 1792MB graphics card 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 added MSAA 4x antialiasing to the standard anisotropic filtering 16x. We enabled antialiasing from the game’s menu. If this was not possible, we forced them using the appropriate driver settings of ATI Catalyst and Nvidia GeForce drivers.
Performance was measured with the games’ own tools and the original demos were recorded if possible. We measured not only the average speed, but also the minimum speed of the cards where possible. Otherwise, the performance was measured manually with Fraps utility version 2.9.8. In the latter case we ran the test three times and took the average of the three for the performance charts.
Starting from version 1.3 we use the game’s integrated benchmarking options together with a custom demo record. Unfortunately, this method does not report the bottom frame rate.
As we might have guessed beforehand, the EVGA card has no advantage over the reference version despite the double amount of onboard memory. Overclocking is far more rewarding as it makes the EVGA comparable to the Radeon HD 4890 and even faster at 2560x1600.
We had some hope about Crysis Warhead, yet this game seems to be quite satisfied with 896 megabytes of local graphics memory, too. The effect from the double amount of memory can only be seen at 2560x1600 but it is small and useless due to the overall low performance of the card.
We disabled the integrated frame rate limiter 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.
Like in the previous test, the EVGA card has a small advantage over Nvidia’s reference sample at 2560x1600 but there is no difference at the other resolutions. Overclocking has the biggest effect at 1920x1200 where it ensures a 20% performance boost in average frame rate. We should remind you that the frame rate limit is set at 30fps in this game for multiplayer mode and a higher frame rate only provides you some reserve for massive battle scenes.
Far Cry 2 is not as demanding an application as its distant relation Crysis, so there is absolutely no difference between the two versions of GeForce GTX 275. However, you can overclock the EVGA card and easily make it as fast as the more expensive GeForce GTX 285. We won’t be surprised if the GeForce GTX 285 will soon share the fate of not-very-lucky solutions from Nvidia such as GeForce 9800 GX2. The single-chip flagship is quite a good device in itself, but it is inappropriately more expensive than the GeForce GTX 275.
Like in the previous tests, there is no difference between the ordinary and 1792MB versions of GeForce GTX 275, but the frequency potential of the EVGA card makes it an appealing buy for overclockers.
The game runs on the Source engine and has an integrated benchmark, but the latter does not report the bottom speed information.
The overclocked EVGA cannot catch up with the GeForce GTX 285 at 2560x1600. Games running on the Source engine usually feature high-quality textures and it is the card’s memory subsystem bandwidth that becomes the decisive factor at such a high resolution. The GeForce GTX 285 has much higher memory bandwidth thanks to the 512-bit memory bus, which explains the defeat of the EVGA card. At the lower resolutions the EVGA is slightly behind the flagship GeForce at the default frequencies and overtakes the latter at the overclocked frequencies.
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 enable the DirectX 10.1 mode for the ATI cards.
The two versions of GeForce GTX 275 go neck and neck, being but slightly different in terms of bottom speed (this may be due to our “manual” method of benchmarking graphics cards in this game). Overclocking does not open new resolutions here: these graphics cards are still limited to 1680x1050.
Like in the other tests, there is no effect from the double amount of memory and a considerable performance boost (22-35% depending on the resolution) from overclocking. We can only note that the non-overclocked GeForce GTX 275 is farther behind the GeForce GTX 285 than in the other tests.
It is only at 2560x1600 that we can see a small effect from the EVGA card’s 1792 megabytes of onboard memory, yet the average frame rate growth is not bigger than 3%. The general picture is like in the previous tests: the increased clock rates ensure a larger advantage than the increased amount of graphics memory.
The larger amount of onboard memory somewhat increases the EVGA’s bottom speed at high resolutions. This does not affect the overall standings, though. Neither the overclocked EVGA nor the GeForce GTX 285 can beat the Radeon HD 4890.
We have to say the same thing again: like other games, Mass Effect is perfectly satisfied with 896 megabytes of onboard graphics memory even at 2560x1600. If you force 4x FSAA on, GeForce GTX 275 class solutions are limited to 1680x1050 and you cannot overcome that barrier by means of ordinary overclocking.
As you can see again, most of today’s games do not need 1 gigabyte of graphics memory. 896 megabytes is quite enough even at the highest resolutions. Perhaps 1GB or more memory may come in handy for GPGPU applications but games are a different story. In this particular case overclocking is useless, too. The GeForce GTX 275 delivers comfortable performance at every resolution up to 2560x1600 even at its default frequencies.
We use the in-game benchmarking tools that do not allow to measure the bottom frame rate. We also enable DirectX 10.1 support for ATI’s solutions.
Although H.A.W.X. is one of the most demanding applications on our list, it finds that 896MB on board the GeForce GTX 275 is enough. The double amount of graphics memory produces no effect. Overclocking, on the contrary, makes the game more enjoyable at 1920x1200 even subjectively. Thus, it is the single case in this review when overclocking provides a tangible benefit for an ordinary gamer.
The game has a frame rate limiter fixed at 30fps.
Nvidia’s solutions are still too slow in Red Alert 3 if you turn FSAA on. Anyway, the EVGA GeForce GTX 275 1792MB behaves exactly like in most of the previous tests. That is, it is no faster than its 896MB cousin.
The double amount of memory on board the EVGA card does not give it any advantage in this game, either. Nvidia’s solutions are limited to 1920x1200 here – the gameplay is too jerky at the higher display modes because the frame rate bottoms out below 20fps.
We could not expect anything new from 3DMark06 which defaults to 1280x1024. And indeed, it is only in the SM3.0/HDR tests that the EVGA GeForce GTX 275 1792MB outperforms the reference card from Nvidia by more than 100 points. When overclocked, it beats the Radeon HD 4890, though.
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’s results are somewhat more of a surprise for us. This benchmark is far more demanding and can make use of all of a graphics card’s resources. However, 1792 megabytes of graphics memory is redundant even here: the results of the two versions of GeForce GTX 275 coincide up to a dozen of points.
It is only in the second test that the EVGA GeForce GTX 275 1792MB has a small advantage over the reference card at 1920x1200, which amounts to a mere 3%. In the other tests these cards have the same result.
As we already noted in our previous reviews, the GeForce GTX 275 is an excellent product that is but slightly inferior to the GeForce GTX 285 while having a much lower price, which makes it far more appealing for the end-user. When it was announced, the only noted drawback was the smaller amount of graphics memory due to the memory bus reduction from 512 to 448 bits. As a result, the new card had 896MB of memory as opposed to the GTX 285’s and the ATI Radeon HD 4890’s 1024MB. We predicted then that there would be versions of the card with GDDR3 chips of double capacity, i.e. with 1792 rather than 896MB of onboard memory. One such model has been tested in this review.
The EVGA GeForce GTX 275 1792MB is neither a revelation nor a disappointment. It is exactly like the ordinary GeForce GTX 275 with 896MB of memory in most games. You can see this in the summary diagrams.
The average advantage of the EVGA card over the reference GeForce GTX 275 is even less than 1% while the maximum gap is only about 2% (at 1920x1200 in Left 4 Dead). Thus, we can say that these cards are identical when it comes to playing modern games. The advantage is larger at 2560x1600, mostly thanks to Crysis Warhead, but the latter game is utterly unplayable on the GeForce GTX 275 irrespective of how much memory it has.
In the other cases, the reference GeForce GTX 275 is no worse than the EVGA card, delivering comfortable or acceptable performance in many games including Call of Duty: World at War, Enemy Territory: Quake Wars, Fallout 3 and Far Cry 2. Our overclocking attempt was far more fruitful: the overclocked EVGA proved to be faster than the GeForce GTX 285 in many games. The average performance gain was 8-12% depending on the resolution while the maximum average performance boost of 35% was achieved in Devil May Cry 4 at 1680x1050 and 1920x1200.
On the other hand, overclocking was really helpful in one case only: it made Tom Clancy’s H.A.W.X. more comfortable to play at 1920x1200.
Thus, EVGA’s product discussed in this review is neither better nor worse than any other GeForce GTX 275 with reference GPU and memory frequencies. What is its positioning then? It is a difficult question, actually. The manufacturer’s site recommends a price of $300 for it, but there is a EVGA GeForce GTX 275 FTW model priced at only $280. The latter has half the amount of memory but its default frequencies are increased from 633/1134 (2268) MHz to 713/1260 (2520) MHz. As we know now, higher clock rates are more important for the card’s performance than an increased amount of memory. Moreover, you can pay $50 more and buy a full-featured GeForce GTX 285 with 1024MB of memory and a 512-bit memory bus.
So, it is unclear who the EVGA GeForce GTX 275 1792MB is meant for if the same manufacturer offers a faster product for less money and a more advanced product for a slightly higher price. The purpose of this product is especially vague because the graphics memory requirements of modern games have reduced due to the popularity of multiplatform projects. 512 megabytes is quite enough for most situations whereas 896 or 1024 megabytes is sufficient for even the most demanding games. We can only suppose that this graphics card may be good for multi-GPU configurations or for CUDA applications.
Summing it up, the EVGA GeForce GTX 275 1792MB is a typical product from EVGA. You may be interested in it if you think that the standard GeForce GTX 275 has too little memory.
Highs:
Lows: