For a little over a year, ATI’s Radeon X800 series GPUs have powered the company’s high-end AGP offerings. Over that time, however, ATI has introduced a number of new graphics products, including the Radeon X850 XT PE that boasts dynamic clock gating and higher clock speeds than previous Platinum Editions, and the Radeon X800 XL built with 0.11-micron fabrication technology. For months, these new graphics cards were limited to PCI Express, but AGP versions of each have recently become available. ATI has also bolstered its AGP lineup with the addition of an All-in-Wonder Radeon X800 XT. Together, these cards breathe new life into a graphics interface that’s been largely ignored for the last six months, but are they any good? Let’s have a look.
The cards
The first and perhaps most interesting addition to ATI’s AGP lineup is the Radeon X800 XL. The X800 XL has already proven itself a potent performer for PCI Express, so we’re happy to see it migrating to AGP. What makes the X800 XL particularly interesting is the fact that its graphics chip is built using 0.11-micron fabrication technology. Well, that and its sixteen pixel pipelines running at 400MHz for about $300, but I digress. You can read more about the Radeon X800 XL’s graphics chip in our review of its PCI Express counterpart.
Physically, the Radeon X800 XL AGP card is huge. At nearly nine inches long, it’s the largest ATI graphics card we’ve ever had in the labs. Note that the card also sports a four-pin Molex connector; a PCI Express x16 slot has enough juice to power the X800 XL on its own, but the AGP version of the card needs a little help.
The AGP version of the X800 XL actually uses a native PCI Express graphics chip in conjunction with an ATI bridge chip translating PCI Express commands to AGP. NVIDIA does something similar with the GeForce 6600 GT AGP, which uses the company’s High Speed Interconnect (HSI) bridge chip as a middleman between the PCI-E graphics chip and AGP interface. The ATI chip runs without a heat sink, though, while NVIDIA’s HSI bridge requires passive cooling.
You’d think ATI would be eager to use its newly developed bridge chip on more graphics cards, but it’s nowhere to be found on the AGP version of the Radeon X850 XT Platinum Edition. ATI hasn’t even snuck the silicon onto the graphics chip’s packaging, suggesting that the company has developed a native AGP version of its R480 GPU.
Like the Radeon X800 XL, the X850 XT Platinum Edition’s GPU serves up a sweet sixteen pixel pipelines, but this time they’re running at 540MHz. That makes this card an absolute fill rate monster, and with GDDR3 running at an effective 1.18GHz, the X850 XT PE has gobs of memory bandwidth, as well. However, the X850 XT PE’s most interesting feature isn’t its peak theoretical performance. The card also has Pentium M-esque dynamic clock gating capabilities, which allows it to turn off inactive portions of the chip. Dynamic clock gating allows the chip to run cooler and consume less power at idle, and when coupled with a temperature-controlled fan, it can also lead to lower noise levels. You can read more about the Radeon X850 XT PE’s graphics chip in our initial review of the X850 series.
Before moving on, we should note that unlike its PCI Express counterpart, the AGP Radeon X850 XT Platinum Edition lacks dual DVI outputs. That’s particularly disappointing considering that the card sells for over $400.
The third entry in our AGP sweepstakes is the All-in-Wonder X800 XT. You won’t find any fancy new GPUs or bridge chips here, just an All-in-Wonder version of ATI’s existing Radeon X800 XT. The AIW X800 XT is currently the fastest member of ATI’s All-in-Wonder lineup, and with NVIDIA’s Personal Cinema all but dead, it doesn’t really have any single-card competition. We’ll only be looking at gaming performance and power consumption today, but stay tuned for more in-depth coverage of the card’s video handling features in an upcoming article.
Our testing methods
All tests were run three times, and their results were averaged, using the following test system.
Processor | Athlon 64 FX-53 2.4GHz | |
System bus | HT 16-bit/1GHz downstream | |
Motherboard | DFI LANParty NF4 Ultra-D | Abit AV8 |
BIOS revision | NF4LD209 | VIA K8T800 Pro |
North bridge | NVIDIA nForce4 Ultra | VIA K8T800 Pro |
South bridge | VIA VT8237 | |
Chipset drivers | ForceWare 6.53 | Hyperion 4.56 |
Memory size | 1GB (2 DIMMs) | 1GB (2 DIMMs) |
Memory type | OCZ PC3200 EL Platinum Rev 2 DDR SDRAM at 400MHz | |
CAS latency (CL) | 2 | 2 |
RAS to CAS delay (tRCD) | 2 | 2 |
RAS precharge (tRP) | 2 | 2 |
Cycle time (tRAS) | 5 | 5 |
Hard drives | Seagate Barracuda 7200.7 NCQ 120GB SATA | |
Audio | nForce4/ALC850 | VT8237/ALC658 |
Audio drivers | Realtek 3.71 | Realtek 3.71 |
Graphics 1 | ATI Radeon X800 XL 256MB with CATALYST 5.4 drivers | ATI Radeon X800 XL 256MB with CATALYST 5.4 drivers |
Graphics 2 | ATI Radeon X800 XL 512MB with CATALYST 5.4 drivers | ATI All-in-Wonder X800 XT 256MB with CATALYST 5.4 drivers |
Graphics 3 | NVIDIA GeForce 6800 GT 256MB with ForceWare 7.89 drivers | ATI Radeon X850 XT PE 256MB with CATALYST 5.4 drivers |
Graphics 4 | NVIDIA GeForce 6800 GT 256MB with ForceWare 7.89 drivers | |
Graphics 5 | NVIDIA GeForce 6800 GT 256MB with ForceWare 7.89 drivers | |
OS | Microsoft Windows XP Professional | |
OS updates | Service Pack 2, DirectX 9.0c |
Today we’re looking at the performance of ATI’s Radeon X800 XL, X850 XT Platinum Edition, and All-in-Wonder X800 XT graphics cards against that of several competitors. On the AGP front, the cards will go toe-to-toe with NVIDIA’s GeForce 6800 GT and Ultra. We’ve also included a handful of PCI Express cards, including a GeForce 6800 GT, and 256 and 512MB flavors of the Radeon X800 XL. If ATI’s bridge chip is doing its job properly, there should be little difference between the performance of AGP and PCI Express versions of the Radeon X800 XL.
Thanks to OCZ for providing us with memory for our testing. If you’re looking to tweak out your system to the max and maybe overclock it a little, OCZ’s RAM is definitely worth considering.
Also, our test system was powered by OCZ PowerStream power supply units. The PowerStream was one of our Editor’s Choice winners in our latest PSU round-up.
We used the following versions of our test applications:
- DOOM 3 1.1 with trdelta1 and trhaze demos
- Far Cry 1.3 with tr1-volcano and tr3-pier demos
- Half-Life 2 with trboat and trtown demos
- Splinter Cell: Chaos Theory 1.02
- FutureMark 3DMark05 Build 120
The test systems’ Windows desktop was set at 1280×1024 in 32-bit color at an 85Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests. All of the 3D gaming tests used the highest possible detail image quality settings except where otherwise noted.
All the tests and methods we employed are publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.
Pixel filling power
Theoretical peak fill rates and memory bandwidth don’t always dictate real world performance, but they should give us an idea of where each card sits relative to its competitors.
Core clock (MHz) | Pixel pipelines | Peak fill rate (Mpixels/s) | Texture units per pixel pipeline | Peak fill rate (Mtexels/s) | Memory clock (MHz) | Memory bus width (bits) | Peak memory bandwidth (GB/s) | |
GeForce 6800 | 325 | 12 | 3900 | 1 | 3900 | 700 | 256 | 22.4 |
GeForce 6600 GT AGP | 500 | 8* | 2000 | 1 | 4000 | 900 | 128 | 14.4 |
GeForce 6600 GT | 500 | 8* | 2000 | 1 | 4000 | 1000 | 128 | 16.0 |
Sapphire Radeon X800 | 392 | 12 | 4704 | 1 | 4704 | 700 | 256 | 22.4 |
Radeon X800 | 400 | 12 | 4800 | 1 | 4800 | 700 | 256 | 22.4 |
GeForce 6800 GT | 350 | 16 | 5600 | 1 | 5600 | 1000 | 256 | 32.0 |
Radeon X800 Pro | 475 | 12 | 5700 | 1 | 5700 | 900 | 256 | 28.8 |
Radeon X800 XL 256MB | 400 | 16 | 6400 | 1 | 6400 | 980 | 256 | 31.4 |
Radeon X800 XL 512MB | 400 | 16 | 6400 | 1 | 6400 | 980 | 256 | 31.4 |
GeForce 6800 Ultra | 425 | 16 | 6800 | 1 | 6800 | 1100 | 256 | 35.2 |
Radeon X800 XT | 500 | 16 | 8000 | 1 | 8000 | 1000 | 256 | 32.0 |
All-in-Wonder X800 XT | 500 | 16 | 8000 | 1 | 8000 | 1000 | 256 | 32.0 |
Radeon X800 XT Platinum Edition | 520 | 16 | 8320 | 1 | 8320 | 1120 | 256 | 35.8 |
Radeon X850 XT | 520 | 16 | 8320 | 1 | 8320 | 1120 | 256 | 35.8 |
Radeon X850 XT Platinum Edition | 540 | 16 | 8640 | 1 | 8640 | 1180 | 256 | 37.8 |
ATI easily wins the peak theoretical fill rate war. The X800 XT and X850 XT Platinum Edition both offer higher fill rates than the GeForce 6800 Ultra, and the Radeon X800 XL boasts higher peak theoretical fill rates than the GeForce 6800 GT. However, the Radeons don’t fare quite so well when we turn our attention to memory bandwidth. Although the Radeon X850 XT Platinum Edition boasts the highest peak theoretical memory bandwidth of the bunch, the All-in-Wonder X800 XT and Radeon X800 XL have less memory bandwidth than the GeForce 6800 Ultra and GT, respectively.
Of course, those are only theoretical peaks. 3DMark05 lets us test fill rates, albeit with synthetic tests.
Despite having a theoretical pixel fill rate advantage over the GeForce 6800 cards, the Radeons can’t match their performance in 3DMark05’s single-texturing fill rate test. It’s been argued that 3DMark05’s single-texturing fill rate test is memory bandwidth-limited, but with the GeForce 6800 Ultra taking top honors ahead of the Radeon X850 XT Platinum Edition, which boasts more memory bandwidth, that may not be the case.
Moving to multi-texturing, the Radeons begin to flex their muscles. Here, the X800 XL has a sizable lead over the GeForce 6800 GT. Heck, the XL almost manages to catch the GeForce 6800 Ultra, whose multi-texturing fill rate trails that of the Radeon X800 XT and X850 XT PE by a significant margin.
In addition to synthetic fill rate tests, 3DMark05 also includes a set of pixel and vertex shader tests.
The Radeons look comparably weak in 3DMark05’s pixel shader test, but bounce back with strong performances in both vertex shader tests.
DOOM 3
Our DOOM 3 tests used the game’s High Detail image quality setting, which enables 8X anisotropic filtering by default.
DOOM 3 has long been an NVIDIA darling, but the Radeon X850 XT PE is only marginally slower than the GeForce 6800 Ultra here. Moving down the line, the All-in-Wonder X800 XT just trails the GeForce 6800 GT, while the Radeon X800 XL falls a little off the pace. Note that AGP and PCI Express variants of the Radeon X800 XL perform nearly identically.
Far Cry
We used “Very high” for all of Far Cry’s in-game quality settings, with the exception of water, which was set to “Ultra high.” In other words, everything was maxed out.
Far Cry reverses the Radeons’ fortunes, and they have no problem keeping the GeForce 6800s at bay. Here, the Radeon X800 XL is left to trade punches with the GeForce 6800 Ultra, while the X800 XT and X850 XT PE race ahead of the pack.
As in DOOM 3, Far Cry shows little performance difference between AGP and PCI Express versions of the Radeon X800 XL.
Half-Life 2
Half-Life 2 was tested with its in-game detail levels turned all the way up. Our benchmark demos explore indoor and outdoor areas from different parts of the game.
The Radeons own Half-Life 2, too. Here, the Radeon X850 XT PE reigns supreme, followed by the All-in-Wonder X800 XT. Again, the Radeon X800 XL is largely competitive with the GeForce 6800 Ultra, at least among 256MB cards.
Interestingly, Half-Life 2 shows some separation between AGP and PCI Express flavors of the Radeon X800 XL, although neither one is a clear winner. Keep in mind that our AGP and PCI Express cards are running on different motherboards with different core logic chipsets, so the platform differences extend beyond the graphics card.
Splinter Cell: Chaos Theory
Splinter Cell: Chaos Theory is the newest addition to our graphics benchmarking suite, and it’s a looker. The game comes with a benchmark demo from the first level, which we’ve used for testing.
Chaos Theory is unique because it supports Shader Model 1.1 and 3.0, but not Shader Model 2.0 or any of its derivatives. The SM 3.0 path offers extra eye candy, including HDR lighting, tone and parallax mapping, and soft shadows that users can enable if they choose. Since the GeForce 6800 cards run a little faster on the SM 3.0 code path, which doesn’t look any worse than the SM 1.1 code path, we’ve used the SM 3.0 path to test them. There’s simply no reason to artificially handicap the 6800 cards’ performance with the SM 1.1 code path. The Radeons were tested with the SM 1.1 code path, which is how end users will have to run the game. Any extra SM 3.0 effects were disabled for the GeForce 6800 cards.
Notch another win for the Radeon X850 XT PE. ATI’s fastest AGP Radeon easily stays ahead of the field in Chaos Theory, while the AIW X800 XT just edges out the GeForce 6800 Ultra for second place. Scores are pretty even between the Radeon X800 XLs and GeForce 6800 GTs.
3DMark05 game tests
Finally, we have 3DMark05’s game tests.
The X850 XT PE comes out ahead again in 3DMark05, where the overall standings closely match our Splinter Cell results.
Power consumption
We busted out our trusty watt meter to measure system power consumption, sans monitor, at the outlet. Power consumption was measured at idle and under a load generated by 3DMark05’s first game test running at 1600×1200 with 4X antialiasing and 16X aniso. Keep in mind that, because our AGP and PCI Express platforms are running with different motherboards and core logic chipsets, differences in system power consumption may not be caused by the graphics card alone.
The Radeon X850 XT PE’s dynamic clock gating helps it to achieve the lowest system power consumption of the lot at idle, while none of the Radeons can match the GeForce 6800 GT’s low power consumption under load. Power consumption is actually pretty even across the board, all things considered.
Conclusions
Performance aside, the most encouraging thing about ATI’s latest crop of AGP Radeons is the fact that all are readily available on the market, often from numerous vendors and board partners. ATI actually announced two of these cards, the Radeon X850 XT Platinum Edition and Radeon X800 XL, at the end of February, so solid current availability isn’t too surprising. Still, it’s nice to be reviewing cards that you can go out and buy, right now.
But which one to buy? At the high end, the Radeon X850 XT Platinum Edition sells for about $460, which isn’t much more expensive than the GeForce 6800 Ultra, which can be had for as little as $420. To NVIDIA’s credit, most versions of the Ultra have dual DVI ports, but the Radeon X850 XT Platinum Edition’s performance is much better across a wider range of games. If you want the fastest AGP card around, the X850 XT PE is the way to go.
Of course, dropping a lot of coin on a top-of-the-line AGP card probably isn’t the best idea considering that the majority of new chipsets are PCI Express-only. Motherboard upgrade options will only become more limited down the road, and if you don’t need the absolute fastest graphics card on the block, the $300 Radeon X800 XL is a pretty sweet deal. The X800 XL is generally faster than the GeForce 6800 GT, and in some cases, it nips at the Ultra’s heels.
And then there’s the All-in-Wonder X800 XT, which is hard to handicap since it really has no peers. Performance-wise, the X800 XT is impressive, but the kicker is its integrated TV tuner and bundled remote. All told, it’s a pretty phenomenal deal for just $360. With higher clock speeds than the X800 XL and its All-in-Wonder extras, the AIW X800 XT is easily the pick of the litter.