Faster And Cheaper
A manufacturing process shrink is a lot to get excited about for both vendors and their customers, as the change normally leads to increased performance and efficiency, while reducing production cost by increasing the number of units produced per wafer. But the eventual benefit to design firms often comes at a huge up-front cost, since significant alterations normally result in bugs that need to be fixed before a full production run can begin. Each test run costs hundreds of thousands of dollars, so it pays to get things right the first or second time. Production delays are an even costlier problem when new architecture is involved, which is why Nvidia normally updates its current products before introducing any new ones.
Today’s updated product, the GeForce GTX 285, makes the typical promises of improved performance and efficiency compared to the GeForce GTX 280 on which it is based. Let’s take a quick look at how it compares to other high-end solutions.
| GeForce GTX 285 | GeForce GTX 280 | GeForce GTX 260 | GeForce GTX 295 | Radeon HD 4870 X2 |
|---|---|---|---|---|---|
Manufacturing Process | 55 nm TSMC | 65 nm TSMC | 65 nm TSMC | 55 nm TSMC | 55 nm TSMC |
SPs | 240 | 240 | 216 | 480 Total | 1,600 |
Core Clock | 648 MHz | 602 MHz | 576 MHz | 576 MHz | 750 MHz |
Shader Clock | 1,476 MHz | 1,296 MHz | 1,242 MHz | 1,242 MHz | 750 MHz |
Memory Data Rate | 2,484 MHz | 2,214 MHz | 1,998 MHz | 1,998 MHz | 3,600 MHz |
Frame Buffer | 1 GB | 1 GB | 896 MB | 1,792 MB Tot. | 2 GB Tot. |
Memory Bus Width | 512-bit | 512-bit | 448-bit | 448-bit x 2 | 256-bit x 2 |
ROPs | 32 | 32 | 28 | 56 Total | 32 Total |
Price | $380 | ~$340 | ~$260 | ~$500 | ~$430 |
The GeForce GTX 285’s most noticeable performance-oriented improvement is an increase in GPU clock speed of around 8%. The memory clock increase–while much larger at 12%–is likely not as important for performance. GeForce GTX 295 graphics units get two of these processors, although each one is handicapped with slower GPU speed, memory speed, and memory bus width. Meanwhile, AMD earns bragging right for both GPU clock and memory data rates, but only because less-complex graphics processors typically clock higher and GDDR5 memory uses a quad-data rate bus.
The specific card in today’s review is a special "XXX" sample of XFX’s GeForce GTX 285, model GX-285N-ZDDF, sporting a core clock of 670 MHz and GDDR3-2500. The larger numbers look more impressive than they are, since these are less than 4% GPU and 1% RAM above the reference specification, so we’ll split the difference and consider it a likely 2-3% average improvement over base speed.
In addition to the basics, XFX includes a door tag, the game Far Cry 2, a DVI-to-HDMI adapter, and an S/PDIF breakout cable. The breakout cable connects a motherboard’s internal S/PDIF audio output to an input adjacent to the card’s power connections, and the combined audio/video signal can be accessed though the output of the HDMI adapter. While this method has been available on Nvidia products for several generations, many previous packages did not include the special cable.
Test Settings
Test System Configuration | |
|---|---|
CPU | Intel Core i7 920 (2.66 GHz, 8.0 MB Cache) |
Overclocked to 4.00 GHz (BCLK 200) | |
CPU Cooler | Swiftech Liquid Cooling: Apogee GTZ water block |
MCP-655b pump, and 3x120 mm radiator | |
Motherboard | Gigabyte EX58-Extreme |
Intel X58/ICH10R Chipset, LGA-1366 | |
RAM | 6.0 GB Crucial DDR3-1600 Triple-Channel Kit |
Overclocked to CAS 8-8-8-16 | |
GTX 285 Graphics | XFX GeForce GTX 285 XXX Edition |
670 MHz GPU, GDDR3-2500 | |
GTX 280 Graphics | EVGA GeForce GTX 280 PN: 01G-P3-1280-AR |
602 MHz GPU, GDDR3-2214 | |
GTX 295 Graphics | GeForce GTX 295 |
2x 576 MHz GPU, GDDR3-1998 | |
Radeon HD 4870 X2 Graphics | Sapphire HD 4870 X2 PN: 100251SR |
2x 750 MHz GPU, GDDR5-3600 | |
Hard Drives | Seagate Barracuda ST3500641AS |
0.5 TB, 7,200 RPM, 16 MB Cache | |
Sound | Integrated HD Audio |
Network | Integrated Gigabit Networking |
Power | Cooler Master RS-850-EMBA |
ATX12V v2.2. EPS12V, 850W, 64A combined +12 V | |
Optical | LG GGC-H20LK 6X Blu-ray/HD DVD-ROM, 16X DVD±R |
Software | |
OS | Microsoft Windows Vista Ultimate x64 SP1 |
Graphics | NVidia Forceware 181.20 Beta |
AMD 8.561.3.0000 Beta | |
Chipset | Intel INF 8.3.0.1016 |
Pooling the hardware from our previous GeForce GTX 295 quad-SLI comparison allows curious readers to compare today’s results to those of more elaborate, multi-card configurations. Highlights of our system include an Intel Core i7 920 processor overclocked to 4.00 GHz, a Swiftech Apogee-GTZ based liquid cooling kit for overclocking stability, and 6 GB of Crucial DDR3-1600 at moderately-tight CAS 8 timings.
Benchmark Configuration | |
|---|---|
Call of Duty: World at War | Patch 1.1, FRAPS/saved game |
Crysis | Patch 1.2.1, DirectX 10, 64-bit executable, benchmark tool |
Far Cry 2 | DirectX 10, Steam Version, in-game benchmark |
Left 4 Dead | Very-High Details, No AA/No AF, vsync off |
World in Conflict | Patch 1009, DirectX 10, timedemo |
3D Mark Vantage | Version 1.02: 3DMark, GPU, CPU scores |
Benchmark Results: COD World At War
We’ve included dual-GPU cards in today’s comparison, but know that this type of comparison is only fair for people considering the more expensive products on a price-per-performance basis. Because most single-GPU graphics cards buyers would not even consider a more expensive dual-GPU solution, we’ve taken the unprecedented step of arranging today’s charts by performance-per-GPU, rather than absolute performance. Absolute performance numbers are still shown, however.
The GeForce GTX 285 edges out the GTX 280 by around 4% in World at War without anti-aliasing
(AA) or anisotropic filtering (AF), with the lead increasing slightly at higher resolutions.
With 4x AA and the game’s AF slider at its highest position, the GeForce GTX 285’s lead over the GTX 280 extends to an average of over 6%. The card posts its greatest gain at the highest test resolution. Both cards offer smooth game play in this title at all settings.
Benchmark Results: Crysis
Even after all this time, Crysis continues to punish graphics cards with high details and minimal optimization. While most of the world is more interested in the more efficient Crysis Warhead version, inclusion of this older title in our GTX 295 Quad-SLI evaluation caused it to reappear in today’s review.
While the test appeared to run smoothly on the GTX 285 at our lowest-tested Crysis settings, occasional stutters would get the player fragged in a real game. The only good option for a single-GPU card would be to reduce detail levels, although the dual-GPU GeForce GTX 295 could make 1920x1200 pixel game play possible for buyers with more discretionary income.
Players can forget about using very high details with AA and AF enabled in Crysis, as even the dual-GPU cards suffered enough stutter to cause an occasional surprise ending. The GTX 285 edged out the GTX 280, but with all cards producing unplayable frame rates, this win is purely academic.
Benchmark Results: Far Cry 2
With highly-detailed maps and textures but unrealistic soft objects, Far Cry 2 puts only a moderate load on modern hardware.
The GeForce GTX 285 maintains its small lead over the GeForce GTX 280, with average performance gains exceeding 6%. Slight stuttering at our highest FC2 test setting hurts the GTX 280.
Benchmark Results: Left 4 Dead
Left 4 Dead is a fun game with loads of gore, but the highest detail level still isn’t very demanding. While the game is playable at its highest settings using an upper midrange card, the test results indicate overall performance differences in high-end parts.
The GeForce GTX 285 leads the GTX 280 by an average of over 7% in Left 4 Dead, but the most noticeable gains are at lower resolutions.
Benchmark Results: World In Conflict
It’s easier to put up with slight stuttering in real-time strategy games (RTS) games as long as the problem is infrequent and the delay is in microseconds. World In Conflict has the detail level needed to reveal such problems.
GeForce GTX 285 performance gains in World in Conflict were unbelievably high with AA and AF disabled, but retesting several times only revealed that the big FPS difference occurred during the in-game benchmark intro when the bombers fly in. Enabling AA and AF brought us back to reality regarding the new card’s true-performance advantage.
Benchmark Results: 3DMark Vantage
Unlike real-world games, 3DMark requires a Physics benchmark run in order to provide a full score. That benchmark runs amazingly slow using the CPU as a physics processor and incredibly fast when using the GPU. Of course, the GPU Physics feature comes in the form of PhysX, which is a proprietary Nvidia technology. 3DMark awards the added "performance points" to the CPU score, since it’s using the GPU as a CPU.
During actual game play, enabling PhysX slows the system slightly, while disabling PhysX disables advanced-physics calculations. Trading a few frames per second (FPS) for increased realism is viable for games, but judging any performance difference between competing products requires that all products support the same setting. Adding proprietary calculations removes standardization, which is a benchmark that’s supposed to be a standardized test, so we chose the Disable PPU option in 3DMark to make this an apples-to-apples comparison.
3DMark and GPU test scores favor the GeForce GTX 285 over the GTX 280 by an average of 7 to 8% percent.
CPU scores are separated by less than half of one percent, which is an expected result when the Disable PPU setting prevents benchmark inflation.
Power And Efficiency
The GeForce GTX 285 replaces the GTX 280, so we used the GTX 280 as the basis for comparing performance-and-efficiency differences.
The GTX 285 we tested today beats the GeForce GTX 280 by almost 9%, but remember that our sample came from XFX with a slight overclock. We noted that the performance gain due to overclocking was likely around 2 to 3% on average, so the typical GeForce GTX 285 performance advantage over the GTX 280 is probably closer to 6%.
The GeForce GTX 280 and our mildly-overclocked GTX 285 have similar peak power consumption, but the GTX 285 saves around 5 W at idle. A non-overclocked GTX 285 would have lower peak power consumption, but the difference is likely only a few watts.
Efficiency is a ratio of energy to performance. Because the GeForce GTX 280 consumed the same peak power as our GTX 285 did, performance improvements translate directly into increased efficiency.
Conclusion
The GeForce GTX 285 consumes less power on average while producing better performance than the GTX 280, and its smaller die allows Nvidia to produce more parts per wafer at a reduced cost. It would appear that everybody wins.
The only problem for consumers is that because the GTX 285 performs better than the GTX 280, Nvidia can charge more for it. An average performance increase of around 6% for the base card (8.75% for our mildly-overclocked test sample) has been met by a price difference of around 10%. It’s up to buyers to decide whether or not they’d like to quibble over a 4% reduction in value, but there’s one thing we can say with certainty: if you already own a GeForce GTX 280, an upgrade to the GTX 285 isn’t practical. GTX 280 owners who want a reasonable performance improvement would be better advised to purchase a second matching card and use the pair in SLI.
Better news for value seekers is that today’s GeForce GTX 285 graphics cards are cheaper than GTX 280 cards were only a few weeks ago. Rather than create a new price class for the GTX 285, Nvidia lowered the price of the GTX 280. SLI upgrades were just given a big value boost for current GTX 280 owners.
But if the GTX 280 is a slightly better value, why would anyone choose the GTX 285? With so small a performance increase, we’d instead consider the card’s improved efficiency. Furthermore, we expect that GTX 280 inventory to be depleted within a few months, making the GTX 285 a better choice for buyers who might be considering an eventual upgrade to SLI.
Noticeable improvements make the GTX 285 a good solution for new systems, but its value as an upgrade part is purely dependent on the inadequacy of the part it will replace.
Source: Tom's Hardware US – Keywords: geforce, gtx, 285
No comments:
Post a Comment