We are a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for us to earn fees by linking to Amazon.com and affiliated sites.
[nextpage title=”Introduction”]
Today AMD finally released a competitor to NVIDIA’s GeForce 8800 family, ATI Radeon HD 2900 XT (formerly known by its codename, R600). We will take a look at the performance of this new release by AMD and compare it to other video cards.
Figure 1: ATI Radeon HD 2900 XT.
The new ATI Radeon HD 2900 XT runs at 740 MHz and accesses its 512 MB GDDR3 memory at 1.65 GHz (825 MHz x 2), using a new 512-bit memory interface, with boosts the memory maximum theoretical transfer rate to 105.6 GB/s – Radeon X1950 XTX has a memory maximum transfer rate of 64 GB/s and GeForce 8800 GTX, of 86.4 GB/s, but the new GeForce 8800 Ultra reaches 103.6 GB/s.
Since it is based on Shader 4.0 (DirectX 10), this video card uses a unified shader processing architecture, i.e., instead of having separated units for processing pixel shader, vertex shader, physics and geometry, it has several generic units, called “stream processors”. This model has 320 processing units, against 128 on GeForce 8800 GTX and Ultra.
There are several other new features on this new video card, like the new generation of Avivo – 2D-enhancement technology –, dubbed Avivo HD, now capable of decoding HD-DVD and Blu-Ray discs directly on the GPU and thus freeing up the CPU from this task, and the use of 128-bit High Dynamic Range (HDR), against 64-bit on the previous ATI generation. We’ve written a full article explaining the architecture of this new video card and we recommend you to read this article to complement the present review.
For a full comparison between ATI Radeon HD 2900 XT and competing chips from NVIDIA, please read our tutorial NVIDIA Chips Comparison Table.
On Figures 2 and 3 you can see the reviewed card from AMD, which is a reference model. When a high-end video card is released, AMD (or NVIDIA) manufactures the cards (actually the cards are manufactured by an Asian manufacturer contracted by them) and then sells them to their partners, the video card manufacturers, which add their sticker, changed the BIOS with their settings, add their cables and CDs and put everything into their box. In this case, the cards provided by different manufacturers are exactly the same. This is the case with Radeon HD 2900 XT, at least right now.
Figure 2: ATI Radeon HD 2900 XT.
Figure 3: ATI Radeon HD 2900 XT, back view.
[nextpage title=”Power Issues”]
The new Radeon HD 2900 XT requires two extra power connectors – as its chip is still manufactured under 80 nm process, it has a typical power consumption of 215 W and AMD recommends at least a 500 W power supply for this baby (750 W if in CrossFire configuration) – however AMD certified some 400 W units to be used with this new video card. A list of certified power supplies can be found here.
The problem is that the second power connector is an 8-pin PCI Express auxiliary power connector specified by the new PCI Express 2.0 standard, and almost all power supplies don’t have this power connector yet. That is why AMD is releasing together with this video card a list of “approved” power supplies, with some manufacturers releasing “special” units for this video card today.
Figure 4: Power connectors on Radeon HD 2900 XT.
To make things worse, this new 8-pin connector resembles a lot the 8-pin power connector used by motherboards, called EPS12V. The problem is that the wires on the 8-pin PCI Express auxiliary power connector is reversed in comparison with the EPS12V connector: where on the first is the ground wires, on the second is the +12 V wires and vice-versa. In other words: if you manage to install the EPS12V connector on you video card you may burn the card, the power and other components (we hope the short-circuit protection from your power supply will prevent this disaster).
In Figure 5 you see an EPS12V compared to a 6-pin PCI Express auxiliary power connector. See how the positions of the black and yellow wires are reversed. Also see how on our EPS12V connector there was a sticker saying “CPU only” (this cable was taken from a Galaxy 1,000 W power supply from Enermax). Not all power supplies come with this sticker or any other warning label.
Figure 5: EPS12V vs. PCI Express auxiliary power connector.
Of course in theory the EPS12V connector doesn’t fit the 8-pin socket on the video card, however we found out that depending on how your EPS12V connector is built this installation is in fact possible. In Figure 6, you can see that we were able to force an EPS12V into the 8-pin auxiliary power connector on our Radeon HD 2900 XT. This was possible because our EPS12V connector was divided into two 4-pin connectors. The power supply used here was a Zalman ZM600-HP.
Figure 6: EPS12V connector installed on the video card. DON’T DO THIS.
If your power supply doesn’t provide this new connector, you may improvise by using two 6-pin PCI Express auxiliary power connectors, as shown in Figure 7.
Figure 7: Improvising by using two 6-pin PCI Express auxiliary power connectors.
[nextpage title=”ATI Radeon HD 2900 XTX”]
We removed the cooler used by this video card and you can see it on Figures 8, 9 and 10. As you can see, it is 100% made of copper using two heat-pipes. The cooler also touches the memory chips.
Figure 8: Cooler used by Radeon HD 2900 XT.
Figure 9: Copper heatsink used by Radeon HD 2900 XT.
Figure 10: Copper base and heatsink used by Radeon HD 2900 XT. See also the two copper heat-pipes.
In Figure 11 you have an overall look of the Radeon HD 2900 XT without its cooler.
Figure 11: Radeon HD 2900 XT without its cooler.
[nextpage title=”ATI Radeon HD 2900 XTX (Cont’d)”]
This video card uses 16 GDDR3 256-Mbit 1 ns chips from Hynix (HY5RS573225AFP-1) as you can see in Figure 12, making the 512 MB of memory this video card has. These chips can run up to 2 GHz (1,000 MHz x 2). Since on this video card the memories were running at 1.65 GHz, there is huge 20% headroom for overclocking the memories inside their specs. Of course you can even try pushing them above their specs.
Figure 12: Hynix GDDR3 256-Mbit 1 ns chip.
On Radeon HD 2900 XT there is a Rage Theater 200 chip, in charge of video capture function of this card (VIVO).
Figure 13: Rage Theater 200.
In Figure 14, you can see the HDMI adapter that comes with the card. One of the new features of the new Radeon HD 2000 family is the built-in support for HDMI (which is a new kind of digital video connector), with support for digital audio on the same connector (several video cards providing HDMI output does not provide digital audio output on the same connector).
Figure 14: HDMI adapter.
[nextpage title=”Main Specifications”]
- Graphics chip: ATI Radeon HD 2900 XT, running at 740 MHz.
- Memory: 512 MB GDDR3 memory (1 ns, 512-bit interface) from Hynix (HY5RS573225AFP-1), running at 1.65 GHz (825 MHz DDR).
- Bus type: PCI Express x16.
- Connectors: Two DVI supporting HDMI and one mini-DIN for S-Video output, Component Video output and video capture (VIVO).
- Video Capture (VIVO): Yes, controlled by Rage Theater 200 chip.
- Number of CDs/DVDs that come with this board: N/A.
- Games that come with this board: N/A.
- Programs that come with this board: None.
- More information: https://ati.amd.com
- Suggested price for the US market: USD 399
[nextpage title=”How We Tested”]
During our benchmarking sessions, we used the configuration listed below. Between our benchmarking sessions the only different device was the video card being tested.
Hardware Configuration
- Motherboard: ASUS P5B (Intel P965, 0904 BIOS)
- CPU: Core 2 Extreme X6800 (dual-core, 2.93 GHz)
- CPU Cooler: Gigabyte Neon 775-BL
- Memory: 2 GB PC-1066/PC2-8500 (Corsair TWIN2X2048-8500C5 kit), configured at 1,066 MHz with 5-5-5-15 timings.
- Hard Drive: Samsung HD080HJ (80 GB, SATA-300, 8 MB buffer, 7,200 rpm)
- Power Supply: Zalman ZM-600HP
- Video Monitor: Samsung SyncMaster 1100MB
- Screen resolution: [email protected] Hz
Software Configuration
- Windows XP Professional installed using NTFS
- Service Pack 2
- Direct X 9.0c
- Intel inf driver version: 8.0.1.1002
- ATI video driver version: Catalyst 8.37 (Radeon HD 2900 XT)
- ATI video driver version: Catalyst 7.2 (other video cards)
- NVIDIA video driver version: 93.71
- NVIDIA video driver version: 97.92 (GeForce 8800 GTS)
Used Software
- 3DMark03 Professional Edition 3.6.0
- 3DMark06 Professional Edition 1.10
- Battlefield 2142 1.01
- Far Cry 1.4 with HardwareOC Far Cry Benchmark 1.7
- F.E.A.R. 1.08
- Quake 4 1.3
We adopted a 3% error margin; thus, differences below 3% cannot be considered relevant. In other words, products with a performance difference below 3% should be considered as having similar performance.
[nextpage title=”3DMark03″]
3DMark03 simulates DirectX 8 and 9 games. Even though this program may be considered “old”, we ran it to see how the tested video cards perform on older games. Since we are comparing high-end cards, we ran this program in higher resolutions, as performance difference between high-end video cards is more expressive on such resolutions. We chose three resolutions, 1600×1200, 1920×1440 and 2048×1536. At each resolution we simulated two scenarios, first with no image quality enhancements enabled (this scenario we called “low”) and then with anti-aliasing set at 4x and anisotropic filtering set at 16x (this scenario we called “high”). The results you check below. All video cards listed below were running with the default clock rates defined by the chip manufacturer.
| 3DMark03 Professional Edition 3.6.0 -1600×1200 | Score | Difference |
| Radeon HD 2900 XT (AMD) | 23150 | |
| GeForce 7950 GX2 (XFX) | 22565 | 2.59% |
| GeForce 8800 GTS 320 MB (MSI) | 15392 | 50.40% |
| GeForce 7950 GT (Zogis) | 14387 | 60.91% |
| GeForce 7900 GT (XFX) | 12020 | 92.60% |
| GeForce 7800 GTX (XFX) | 11744 | 97.12% |
| Radeon X1950 Pro (PowerColor) | 11243 | 105.91% |
| GeForce 7800 GT (NVIDIA) | 10278 | 125.24% |
| 3DMark03 Professional Edition 3.6.0 -1920×1440 | Score | Difference |
| GeForce 7950 GX2 (XFX) | 18351 | 3.75% |
| Radeon HD 2900 XT (AMD) | 17687 | |
| GeForce 8800 GTS 320 MB (MSI) | 12081 | 46.40% |
| GeForce 7950 GT (Zogis) | 11287 | 56.70% |
| GeForce 7900 GT (XFX) | 9437 | 87.42% |
| GeForce 7800 GTX (XFX) | 9245 | 91.31% |
| Radeon X1950 Pro (PowerColor) | 8731 | 102.58% |
| GeForce 7800 GT (NVIDIA) | 8018 | 120.59% |
| 3DMark03 Professional Edition 3.6.0 – 2048×1536 | Score | Difference |
| GeForce 7950 GX2 (XFX) | 16953 | 4.47% |
| Radeon HD 2900 XT (AMD) | 16228 | |
| GeForce 8800 GTS 320 MB (MSI) | 11327 | 43.27% |
| GeForce 7950 GT (Zogis) | 10348 | 56.82% |
| GeForce 7800 GTX (XFX) | 8386 | 93.51% |
| Radeon X1950 Pro (PowerColor) | 7656 | 111.96% |
| GeForce 7800 GT (NVIDIA) | 7272 | 123.16% |
| GeForce 7900 GT (XFX) | 7066 | 129.66% |
| 3DMark03 Professional Edition 3.6.0 – 1600×1200 AAx4, AFx16 | Score | Difference |
| GeForce 7950 GX2 (XFX) | 13181 | 29.25% |
| Radeon HD 2900 XT (AMD) | 10198 | |
| GeForce 8800 GTS 320 MB (MSI) | 8637 | 18.07% |
| GeForce 7950 GT (Zogis) | 7903 | 29.04% |
| GeForce 7900 GT (XFX) | 6634 | 53.72% |
| GeForce 7800 GTX (XFX) | 6422 | 58.80% |
| Radeon X1950 Pro (PowerColor) | 6236 | 63.53% |
| GeForce 7800 GT (NVIDIA) | 5553 | 83.65% |
| 3DMark03 Professional Edition 3.6.0 – 1920×1440 AAx4, AFx16 | Score | Difference |
| GeForce 7950 GX2 (XFX) | 10231 | 30.95% |
| Radeon HD 2900 XT (AMD) | 7813 | |
| GeForce 8800 GTS 320 MB (MSI) | 6535 | 19.56% |
| GeForce 7950 GT (Zogis) | 6037 | 29.42% |
| Radeon X1950 Pro (PowerColor) | 4765 | 63.97% |
| GeForce 7900 GT (XFX) | 4637 | 68.49% |
| GeForce 7800 GTX (XFX) | 4467 | 74.90% |
| GeForce 7800 GT (NVIDIA) | 3850 | 102.94% |
| 3DMark03 Professional Edition 3.6.0 – 2048×1536 AAx4, AFx16 | Score | Difference |
| GeForce 7950 GX2 (XFX) | 9235 | 29.91% |
| Radeon HD 2900 XT (AMD) | 7109 | |
| GeForce 8800 GTS 320 MB (MSI) | 5494 | 29.40% |
| GeForce 7950 GT (Zogis) | 5409 | 31.43% |
| GeForce 7900 GT (XFX) | 4187 | 69.79% |
| Radeon X1950 Pro (PowerColor) | 4091 | 73.77% |
| GeForce 7800 GTX (XFX) | 4030 | 76.40% |
| GeForce 7800 GT (NVIDIA) | 3470 | 104.87% |
[nextpage title=”3DMark06″]
3DMark06 simulates DirectX 9.0c (Shader 3.0) games and it also puts HDR (High Dynamic Range) into the equation to calculate its final score. So it simulates the most high-end games available today. Since we were comparing high-end cards, we ran this program in higher resolutions, as performance difference between high-end video cards is more expressive on such resolutions. We chose three resolutions, 1600×1200, 1920×1440 and 2048×1536. The results you check below. All video cards listed below were running with the default clock rates defined by the chip manufacturer.
| 3DMark06 Professional Edition 1.10 – 1600×1200 | Score | Difference |
| Radeon HD 2900 XT (AMD) | 9448 | |
| GeForce 8800 GTS 320 MB (MSI) | 7504 | 25.91% |
| GeForce 7950 GX2 (XFX) | 7208 | 31.08% |
| GeForce 7950 GT (Zogis) | 4679 | 101.92% |
| Radeon X1950 Pro (PowerColor) | 4504 | 109.77% |
| GeForce 7900 GT (XFX) | 3937 | 139.98% |
| GeForce 7800 GTX (XFX) | 3861 | 144.70% |
| GeForce 7800 GT (NVIDIA) | 3311 | 185.35% |
| 3DMark06 Professional Edition 1.10 – 1920×1440 | Score | Difference |
| Radeon HD 2900 XT (AMD) | 7982 | |
| GeForce 8800 GTS 320 MB (MSI) | 6414 | 24.45% |
| GeForce 7950 GX2 (XFX) | 6074 | 31.41% |
| GeForce 7950 GT (Zogis) | 3876 | 105.93% |
| Radeon X1950 Pro (PowerColor) | 3744 | 113.19% |
| GeForce 7900 GT (XFX) | 3187 | 150.45% |
| GeForce 7800 GTX (XFX) | 3117 | 156.08% |
| GeForce 7800 GT (NVIDIA) | 2657 | 200.41% |
| 3DMark06 Professional Edition 1.10 – 2048×1536 | Score | Difference |
| Radeon HD 2900 XT (AMD) | 7487 | |
| GeForce 8800 GTS 320 MB (MSI) | 6035 | 24.06% |
| GeForce 7950 GX2 (XFX) | 5686 | 31.67% |
| GeForce 7950 GT (Zogis) | 3603 | 107.80% |
| Radeon X1950 Pro (PowerColor) | 3427 | 118.47% |
| GeForce 7900 GT (XFX) | 2949 | 153.88% |
| GeForce 7800 GTX (XFX) | 2882 | 159.78% |
| GeForce 7800 GT (NVIDIA) | 2450 | 205.59% |
[nextpage title=”Quake 4″]
Quake 4 uses the same game engine as Doom 3 and, since we are comparing high-end cards, we ran this program in higher resolutions, as performance difference between high-end video cards is more expressive on such resolutions. We chose two resolutions, 1600×1200 and 2048×1536, first with image quality set at “low” and then with image quality set at “high”. We upgraded this game to version 1.3 and run the id_demo001 net demo that comes with this version. Click here for more details on how to use Quake 4 to benchmark a system. The results you check below and are given in frames per second. All video cards listed below were running with the default clock rates defined by the chip manufacturer.
| Quake 4 1.3 – 1600×1200 – Low | Score | Difference |
| GeForce 8800 GTS 320 MB (MSI) | 123.46 | 13.69% |
| GeForce 7950 GX2 (XFX) | 116.70 | 7.47% |
| Radeon HD 2900 XT (AMD) | 108.59 | |
| GeForce 7950 GT (Zogis) | 99.87 | 8.73% |
| GeForce 7900 GT (XFX) | 84.35 | 28.74% |
| GeForce 7800 GTX (XFX) | 83.15 | 30.60% |
| GeForce 7800 GT (NVIDIA) | 73.48 | 47.78% |
| Radeon X1950 Pro (PowerColor) | 68.01 | 59.67% |
| Quake 4 1.3 – 2048×1536 – Low | Score | Difference |
| GeForce 7950 GX2 (XFX) | 106.07 | 4.47% |
| Radeon HD 2900 XT (AMD) | 101.53 | |
| GeForce 8800 GTS 320 MB (MSI) | 99.90 | 1.63% |
| GeForce 7950 GT (Zogis) | 70.20 | 44.63% |
| GeForce 7900 GT (XFX) | 58.32 | 74.09% |
| GeForce 7800 GTX (XFX) | 57.29 | 77.22% |
| GeForce 7800 GT (NVIDIA) | 50.27 | 101.97% |
| Radeon X1950 Pro (PowerColor) | 41.14 | 146.79% |
| Quake 4 1.3 – 1600×1200 – High | Score | Difference |
| GeForce 8800 GTS 320 MB (MSI) | 119.54 | 14.50% |
| GeForce 7950 GX2 (XFX) | 113.26 | 8.49% |
| Radeon HD 2900 XT (AMD) | 104.40 | |
| GeForce 7950 GT (Zogis) | 96.33 | 8.38% |
| GeForce 7900 GT (XFX) | 81.33 | 28.37% |
| GeForce 7800 GTX (XFX) | 80.17 | 30.22% |
| GeForce 7800 GT (NVIDIA) | 70.91 | 47.23% |
| Radeon X1950 Pro (PowerColor) | 63.37 | 64.75% |
| Quake 4 1.3 – 2048×1536 – High | Score | Difference |
| GeForce 7950 GX2 (XFX) | 102.71 | 8.14% |
| GeForce 8800 GTS 320 MB (MSI) | 96.08 | 1.16% |
| Radeon HD 2900 XT (AMD) | 94.98 | |
| GeForce 7950 GT (Zogis) | 67.80 | 40.09% |
| GeForce 7900 GT (XFX) | 56.44 | 68.28% |
| GeForce 7800 GTX (XFX) | 55.44 | 71.32% |
| GeForce 7800 GT (NVIDIA) | 48.51 | 95.79% |
| Radeon X1950 Pro (PowerColor) | 39.57 | 140.03% |
[nextpage title=”F.E.A.R.”]
F.E.A.R. is a heavy game and we used its internal benchmarking module. We upgraded it to version 1.08 and since we are comparing high-end cards, we ran this program in higher resolutions, as performance difference between high-end video cards is more expressive on such resolutions. We chose three resolutions, 1600×1200, 1920×1440 and 2048×1536. We set “computer settings” to “maximum” and then ran each resolution in two scenarios, first with “graphics card” set at “low” and then with this item set at “maximum”. Let’s take a look at the results, given in frames per second. All video cards listed below were running with the default clock rates defined by the chip manufacturer.
| F.E.A.R. 1.08 – 1600×1200 – Low Quality | Score | Difference |
| Radeon HD 2900 XT (AMD) | 368 | |
| GeForce 7950 GX2 (XFX) | 284 | 29.58% |
| GeForce 8800 GTS 320 MB (MSI) | 237 | 55.27% |
| Radeon X1950 Pro (PowerColor) | 191 | 92.67% |
| GeForce 7950 GT (Zogis) | 168 | 119.05% |
| GeForce 7900 GT (XFX) | 140 | 162.86% |
| GeForce 7800 GTX (XFX) | 135 | 172.59% |
| GeForce 7800 GT (NVIDIA) | 109 | 237.61% |
| F.E.A.R. 1.08 – 1920×1440 – Low Quality | Score | Difference |
| Radeon HD 2900 XT (AMD) | 270 | |
| GeForce 7950 GX2 (XFX) | 215 | 25.58% |
| GeForce 8800 GTS 320 MB (MSI) | 175 | 54.29% |
| Radeon X1950 Pro (PowerColor) | 139 | 94.24% |
| GeForce 7950 GT (Zogis) | 124 | 117.74% |
| GeForce 7900 GT (XFX) | 102 | 164.71% |
| GeForce 7800 GTX (XFX) | 98 | 175.51% |
| GeForce 7800 GT (NVIDIA) | 80 | 237.50% |
| F.E.A.R. 1.08 – 2048×1536 – Low Quality | Score | Difference |
| Radeon HD 2900 XT (AMD) | 242 | |
| GeForce 7950 GX2 (XFX) | 192 | 26.04% |
| GeForce 8800 GTS 320 MB (MSI) | 158 | 53.16% |
| Radeon X1950 Pro (PowerColor) | 122 | 98.36% |
| GeForce 7950 GT (Zogis) | 111 | 118.02% |
| GeForce 7900 GT (XFX) | 91 | 165.93% |
| GeForce 7800 GTX (XFX) | 87 | 178.16% |
| GeForce 7800 GT (NVIDIA) | 70 | 245.71% |
| F.E.A.R. 1.08 – 1600×1200 – Maximum Quality | Score | Difference |
| GeForce 7950 GX2 (XFX) | 65 | 1.56% |
| Radeon HD 2900 XT (AMD) | 64 | |
| GeForce 8800 GTS 320 MB (MSI) | 45 | 42.22% |
| GeForce 7950 GT (Zogis) | 42 | 52.38% |
| Radeon X1950 Pro (PowerColor) | 37 | 72.97% |
| GeForce 7900 GT (XFX) | 36 | 77.78% |
| GeForce 7800 GTX (XFX) | 34 | 88.24% |
| GeForce 7800 GT (NVIDIA) | 30 | 113.33% |
| F.E.A.R. 1.08 – 1920×1440 – Maximum Quality | Score | Difference |
| GeForce 7950 GX2 (XFX) | 48 | 2.13% |
| Radeon HD 2900 XT (AMD) | 47 | |
| GeForce 8800 GTS 320 MB (MSI) | 31 | 51.61% |
| GeForce 7950 GT (Zogis) | 31 | 51.61% |
| Radeon X1950 Pro (PowerColor) | 28 | 67.86% |
| GeForce 7900 GT (XFX) | 25 | 88.00% |
| GeForce 7800 GTX (XFX) | 24 | 95.83% |
| GeForce 7800 GT (NVIDIA) | 21 | 123.81% |
| F.E.A.R. 1.08 – 2048×1536 – Maximum Quality | Score | Difference |
| GeForce 7950 GX2 (XFX) | 42 | 0.00% |
| Radeon HD 2900 XT (AMD) | 42 | |
| GeForce 7950 GT (Zogis) | 28 | 50.00% |
| GeForce 8800 GTS 320 MB (MSI) | 27 | 55.56% |
| GeForce 7900 GT (XFX) | 23 | 82.61% |
| Radeon X1950 Pro (PowerColor) | 22 | 90.91% |
| GeForce 7800 GTX (XFX) | 22 | 90.91% |
| GeForce 7800 GT (NVIDIA) | 17 | 147.06% |
[nextpage title=”Far Cry”]
Far Cry is a heavy game based on the Shader 3.0 (DirectX 9.0c) programming model. We’ve updated the game to version 1.4. To measure the performance we run four times the demo created by German magazine PC Games Hardware (PCGH) and the results presented below are an arithmetic average of the collected data. We used the HardwareOC Far Cry Benchmark 1.7 utility to help us collecting data.
Since we are comparing high-end cards, we ran this game in higher resolutions, as performance difference between high-end video cards is more expressive on such resolutions. We chose three resolutions, 1600×1200, 1920×1440 and 2048×1536. At each resolution we simulated two scenarios, first with no image quality enhancements enabled and graphics details set to “maximum” (this scenario we called “low”) and then with anti-aliasing set at 8x, anisotropic filtering set at 16x and graphics details set to “ultra” (this scenario we called “high”). On all scenarios we set the rendering engine to Shader 3.0. The results, given in frames per second, you check below. All video cards listed below were running with the default clock rates defined by the chip manufacturer.
| Far Cry 1.4 – 1600×1200 – Maximum Details | Score | Difference |
| GeForce 8800 GTS 320 MB (MSI) | 144.75 | 0.23% |
| Radeon HD 2900 XT (AMD) | 144.42 | |
| GeForce 7950 GX2 (XFX) | 143.59 | 0.58% |
| GeForce 7950 GT (Zogis) | 127.31 | 13.44% |
| Radeon X1950 Pro (PowerColor) | 112.27 | 28.64% |
| GeForce 7900 GT (XFX) | 108.81 | 32.73% |
| GeForce 7800 GTX (XFX) | 107.90 | 33.85% |
| GeForce 7800 GT (NVIDIA) | 91.98 | 57.01% |
| Far Cry 1.4 – 1920×1440 – Maximum Details | Score | Difference |
| GeForce 7950 GX2 (XFX) | 138.05 | 6.42% |
| Radeon HD 2900 XT (AMD) | 129.72 | |
| GeForce 8800 GTS 320 MB (MSI) | 120.16 | 7.96% |
| GeForce 7950 GT (Zogis) | 99.27 | 30.67% |
| Radeon X1950 Pro (PowerColor) | 86.91 | 49.26% |
| GeForce 7900 GT (XFX) | 82.73 | 56.80% |
| GeForce 7800 GTX (XFX) | 80.62 | 60.90% |
| GeForce 7800 GT (NVIDIA) | 69.26 | 87.29% |
| Far Cry 1.4 – 2048×1536 – Maximum Details | Score | Difference |
| GeForce 7950 GX2 (XFX) | 132.31 | 8.37% |
| Radeon HD 2900 XT (AMD) | 122.09 | |
| GeForce 8800 GTS 320 MB (MSI) | 108.58 | 12.44% |
| GeForce 7950 GT (Zogis) | 89.44 | 36.50% |
| Radeon X1950 Pro (PowerColor) | 74.99 | 62.81% |
| GeForce 7900 GT (XFX) | 74.45 | 63.99% |
| GeForce 7800 GTX (XFX) | 72.60 | 68.17% |
| GeForce 7800 GT (NVIDIA) | 62.17 | 96.38% |
| Far Cry 1.4 – 1600×1200 – AAx4, AFx16, Ultra Details | Score | Difference |
| GeForce 7950 GX2 (XFX) | 133.22 | 106.86% |
| GeForce 7950 GT (Zogis) | 96.58 | 49.97% |
| GeForce 7900 GT (XFX) | 81.60 | 26.71% |
| GeForce 7800 GTX (XFX) | 78.61 | 22.07% |
| Radeon X1950 Pro (PowerColor) | 75.58 | 17.36% |
| GeForce 7800 GT (NVIDIA) | 67.41 | 4.67% |
| Radeon HD 2900 XT (AMD) | 64.40 | |
| GeForce 8800 GTS 320 MB (MSI) | 62.85 | 2.47% |
| Far Cry 1.4 – 1920×1440 – AAx4, AFx16, Ultra Details | Score | Difference |
| GeForce 7950 GX2 (XFX) | 109.73 | 120.83% |
| GeForce 7950 GT (Zogis) | 72.53 | 45.96% |
| GeForce 7900 GT (XFX) | 60.91 | 22.58% |
| GeForce 7800 GTX (XFX) | 58.93 | 18.60% |
| Radeon X1950 Pro (PowerColor) | 57.07 | 14.85% |
| GeForce 7800 GT (NVIDIA) | 50.04 | 0.70% |
| Radeon HD 2900 XT (AMD) | 49.69 | |
| GeForce 8800 GTS 320 MB (MSI) | 38.58 | 28.80% |
| Far Cry 1.4 – 2048×1536 – AAx4, AFx16, Ultra Details | Score | Difference |
| GeForce 7950 GX2 (XFX) | 100.70 | 122.74% |
| GeForce 7950 GT (Zogis) | 65.46 | 44.79% |
| GeForce 7900 GT (XFX) | 54.82 | 21.26% |
| GeForce 7800 GTX (XFX) | 52.86 | 16.92% |
| Radeon X1950 Pro (PowerColor) | 49.77 | 10.09% |
| Radeon HD 2900 XT (AMD) | 45.21 | |
| GeForce 7800 GT (NVIDIA) | 44.99 | 0.49% |
| GeForce 8800 GTS 320 MB (MSI) | 29.25 | 54.56% |
[nextpage title=”Battlefield 2142″]
Battlefield 2142 is the latest member of the Battlefield franchise. We updated this game to version 1.01. We created our own demo based on Sidi Power Plant map (click here to download the demo we created for this test), which provided a very consistent number of frames per second. We ran it and measured performance with FRAPS. Click here to read in details how we benchmarked using Battlefield 2142.
Since we are comparing high-end cards, we ran this game in higher resolutions, as performance difference between high-end video cards is more expressive on such resolutions. We chose three resolutions, 1600×1200, 1920×1440 and 2048×1536. First we ran our demo with image quality set at “low” (with texture manually set at its minimum level) and then with image quality set at “high” (with anti-aliasing manually set at 4x). Below you can see the results, given in frames per second. All video cards listed below were running with the default clock rates defined by the chip manufacturer.
| Battlefield 2142 1.01 – 1600×1200 – Low | Score | Difference |
| Radeon HD 2900 XT (AMD) | 416.10 | |
| GeForce 7950 GX2 (XFX) | 400.27 | 3.95% |
| GeForce 8800 GTS 320 MB (MSI) | 317.54 | 31.04% |
| GeForce 7950 GT (Zogis) | 233.98 | 77.84% |
| GeForce 7900 GT (XFX) | 195.97 | 112.33% |
| GeForce 7800 GTX (XFX) | 191.74 | 117.01% |
| Radeon X1950 Pro (PowerColor) | 184.70 | 125.28% |
| GeForce 7800 GT (NVIDIA) | 166.70 | 149.61% |
| Battlefield 2142 1.01 – 1920×1440 – Low | Score | Difference |
| Radeon HD 2900 XT (AMD) | 302.05 | |
| GeForce 7950 GX2 (XFX) | 300.40 | 0.55% |
| GeForce 8800 GTS 320 MB (MSI) | 238.76 | 26.51% |
| GeForce 7950 GT (Zogis) | 175.23 | 72.37% |
| GeForce 7900 GT (XFX) | 146.36 | 106.37% |
| GeForce 7800 GTX (XFX) | 142.60 | 111.82% |
| Radeon X1950 Pro (PowerColor) | 140.91 | 114.36% |
| GeForce 7800 GT (NVIDIA) | 123.03 | 145.51% |
| Battlefield 2142 1.01 – 2048×1536 – Low | Score | Difference |
| Radeon HD 2900 XT (AMD) | 271.71 | |
| GeForce 7950 GX2 (XFX) | 271.20 | 0.19% |
| GeForce 8800 GTS 320 MB (MSI) | 214.35 | 26.76% |
| GeForce 7950 GT (Zogis) | 157.38 | 72.65% |
| GeForce 7900 GT (XFX) | 132.57 | 104.96% |
| GeForce 7800 GTX (XFX) | 131.24 | 107.03% |
| Radeon X1950 Pro (PowerColor) | 121.72 | 123.23% |
| GeForce 7800 GT (NVIDIA) | 108.96 | 149.37% |
| Battlefield 2142 1.01 – 1600×1200 – High | Score | Difference |
| GeForce 7950 GX2 (XFX) | 104.83 | 29.53% |
| GeForce 8800 GTS 320 MB (MSI) | 81.33 | 0.49% |
| Radeon HD 2900 XT (AMD) | 80.93 | |
| GeForce 7950 GT (Zogis) | 62.04 | 30.45% |
| Radeon X1950 Pro (PowerColor) | 52.00 | 55.63% |
| GeForce 7900 GT (XFX) | 50.53 | 60.16% |
| GeForce 7800 GTX (XFX) | 48.17 | 68.01% |
| GeForce 7800 GT (NVIDIA) | 42.01 | 92.64% |
| Battlefield 2142 1.01 – 1920×1440 – High | Score | Difference |
| GeForce 7950 GX2 (XFX) | 77.80 | 26.83% |
| Radeon HD 2900 XT (AMD) | 61.34 | |
| GeForce 8800 GTS 320 MB (MSI) | 52.51 | 16.82% |
| GeForce 7950 GT (Zogis) | 45.92 | 33.58% |
| Radeon X1950 Pro (PowerColor) | 40.76 | 50.49% |
| GeForce 7900 GT (XFX) | 36.88 | 66.32% |
| GeForce 7800 GTX (XFX) | 35.05 | 75.01% |
| GeForce 7800 GT (NVIDIA) | 29.85 | 105.49% |
| Battlefield 2142 1.01 – 2048×1536 – High | Score | Difference |
| GeForce 7950 GX2 (XFX) | 69.94 | 27.26% |
| Radeon HD 2900 XT (AMD) | 54.96 | |
| GeForce 8800 GTS 320 MB (MSI) | 46.16 | 19.06% |
| GeForce 7950 GT (Zogis) | 40.79 | 34.74% |
| Radeon X1950 Pro (PowerColor) | 35.83 | 53.39% |
| GeForce 7900 GT (XFX) | 33.15 | 65.79% |
| GeForce 7800 GTX (XFX) | 31.42 | 74.92% |
| GeForce 7800 GT (NVIDIA) | 26.54 | 107.08% |
[nextpage title=”Conclusions”]
Even though our review methodology had two flaws – we didn’t get a GeForce 8800 GTX or a GeForce 8800 Ultra from NVIDIA for comparison and also we didn’t run any DirectX 10 benchmark – we found some interesting results.
Let’s recap what we found out.
On 3DMark03, which simulates older games based on DirectX 9 (Shader 2.0 model), Radeon HD 2900 XT with no image quality settings enabled achieved the same performance level as GeForce 7950 GX2 and was 43-50% faster than GeForce 8800 GTS and 56-61% faster than GeForce 7950 GT. However, when we increased image quality settings, GeForce 7950 GX2 was 30% faster, but Radeon HD 2900 XT was 43-50% faster than GeForce 8800 GTS and 56-61% faster than GeForce 7950GT.
On 3DMark06, which simulates games based on DirectX 9.0c (Shader 3.0 model) and puts a lot of weight on High Dynamic Range (HDR), Radeon HD 2900 XT was 31% faster than GeForce 7950 GX2, 24-25% faster than GeForce 8800 GTS and 100-107% faster than GeForce 7950 GT.
Quake 4 continues to be AMD/ATI’s Achilles’ Heel, even thought Radeon HD 2900 XT improved a little bit on this game: it achieved the same performance level as GeForce 8800 GTS at 2048×1536. At 1600×1200 GeForce 7950 GX2 was 7-8% faster and GeForce 8800 GTS was 14% faster.
On F.E.A.R. with no image quality settings enabled, Radeon HD 2900 XT was 25-30% faster than GeForce 7950 GX2, 53-55% faster than GeForce 8800 GTS and 117-119% faster than GeForce 7950 GT. However when we maxed out the image quality settings, the new video card from AMD achieve the same performance level as GeForce 7950 GX2, being 42-55% faster than GeForce 8800 GTS and 50-52% faster than GeForce 7950 GT.
On Far Cry with no image quality enhancements enabled, Radeon HD 2900 XT achieved the same performance level as GeForce 7950 GX2 and GeForce 8800 GTS at 1600×1200, being 13% faster than GeForce 7950 GT on this same resolution. At the other two resolutions we tested, GeForce 7950 GX2 was 6-8% faster, but the new video card from AMD was 8-12% faster than GeForce 8800 GTS and 30-36% faster than GeForce 7950 GT.
When we enabled image quality settings, Radeon HD 2900 XT achieved the same performance level as GeForce 8800 GTS at 1600×1200 but was 29-54% faster on the other two resolutions we tested. Here, however, all other video cards we included in our comparison were faster than Radeon HD 2900 XT. GeForce 7950 GT was, for example, 45-50% faster.
And finally on Battlefield 2142 with no image quality settings enabled, Radeon HD 2900 XT achieved the same performance as GeForce 7950 GX2, was 26-31% faster than GeForce 8800 GTS and 72-77% faster than GeForce 7950 GT. When we maxed out the image quality settings, GeForce 7950 GX2 was 26-29% faster than Radeon HD 2900 XT, while the new video card from AMD achieved the same performance as GeForce 8800 GTS at 1600×1200 but was 16-19% faster on the other two resolutions we tested, and was also 30-34% faster than GeForce 7950 GT.
Is Radeon HD 2900 XT a good high-end video? Yes, especially for its price range: with a maximum suggested price of USD 399 for the North-American market it makes a terrific option for those who want more performance than a GeForce 8800 GTS can deliver and don’t want to pay for a GeForce 8800 GTX.
However, we got this feeling that AMD is late for the party. As we got close to the Radeon HD 2900 XT launch date, NVIDIA released GeForce 8800 Ultra, leaving NVIDIA once again ahead of the game. And we think that when AMD comes with a competitor to this new NVIDIA threat, NVIDIA will probably come out with a new product line.
It is not only that. Radeon HD 2900 XT is the only DirectX 10 product from AMD, while NVIDIA has currently a complete line of DirectX 10-based video cards, and competing products from AMD are expected to reach the market only by the end of June – giving NVIDIA plenty of time to do whatever they want, including launching more models.
Whether this huge delay is a reflex of the merger between AMD and ATI is still a mystery to us.
Also, we can’t help thinking about GeForce 7950 GX2. This video card was released one year ago and when we maxed out image quality settings it achieved a superior performance on several games – this video card isn’t DirectX 10 and its HDR capability is really limited compared to newer products, though.