The cheapskates of the world seemed to enjoy our recent look at relatively inexpensive video cards, and rightly so. For perhaps the first time in recent memory, we found considerable value in graphics cards that cost less than $100. But today we separate the bargain-minded from the merely cheap by considering a pair of high-end graphics cards that offer substantial value in their own right. Tight competition between AMD and Nvidia has resulted in two new video cards that redefine their end of the market for just a smidgen under 300 bucks. Rather stealthily, Nvidia has ramped up the performance of its GeForce GTX 260 by enabling an additional thread processing cluster and cranking up clock speeds. Meanwhile, AMD has made its Radeon HD 4870 even more potent by doubling the onboard memory to a full gigabyte. The question is: Has Nvidia done enough to catch up with AMD, or will the additional memory allow the Radeon to hold on to its performance advantage? We aim to find out.
The GeForce GTX 260reloaded
As you may know if you’re a hopeless geek/regular reader, the GeForce GTX series of graphics cards is based on Nvidia’s GT200 graphics processor. The GeForce GTX 280 is the full-on version of the GT200, while the less expensive GTX 260 has two of its ten thread processing clusters and one of its eight ROP partitions disabled. Disabling parts of a chip for product segmentation purposes is a common practice, and it can provide a fitting home for chips with portions that are less than perfect, so this basic plan makes sense. Trouble is, Nvidia didn’t count on the Radeon HD 4870 outgunning the GTX 260 at a lower price, which is just what the 4870 did upon its debut. Nvidia has responded in several ways, first by cutting prices and, more recently, but quietly changing the spec on the GTX 260 so that nine of its ten thread processing clusters (TPCs) are enabled.
The additional cluster gives the GTX 260 a little more graphics power, raising its stream processor count from 192 to 216 and its peak texture filtering capacity from 64 texels per clock to 72. In addition to that, prevailing clock speeds are up quite a bit. Let me give you a couple of examples.
The handsomely stickered card you see above is the GeForce GTX 260 AMP²! Edition from Zotac. (I would like to thank Zotac for making me type AMP²!, since I could use the exercise.) Not only does it have an additional TPC, but its clock speeds are quite a bit higher than early GTX 260s. The AMP²! has a 649MHz GPU core, 1404MHz shaders, and 896MB of GDDR3 memory at 1053MHz, up from 576/1242/999MHz on the first wave of GTX 260 cards. Those clock speeds are also, I should note, higher than the stock clocks for the GeForce GTX 280, which are 602/1296/1107MHz.
What this means, essentially, is that the revamped GTX 260 now offers almost all of the performance of the original GTX 280 at under $300, if you count the Zotac AMP²!’s $299.99 price at Newegg as “under $300.” Happily, there aren’t any mail-in rebates at all involved in that price, and Zotac even throws in a copy of the excellent Race Driver GRID. What’s not to like?
Well, maybe one thing. Why the devil did Nvidia hang on to the GTX 260 name when they decided to upgrade their product this substantially? I have it on good authority that the number “270” was available to them, conveniently located between 260 and 280. The mystery of product naming deepens and becomes more opaque with each passing day.
But, hey, faster games and stuff, so I’ll get over it. I’m just going to call the version with 216 SPs the GeForce GTX 260 Reloaded for clarity’s sake.
Zotac covers the AMP²! with a limited lifetime warranty, but you’ve gotta register via Zotac’s website within 30 days of purchase or the warranty drops to a two-year term. That’s the deal for North America, at least; other regions are different. Also, confusingly, going to Zotac’s main website and telling it you’re from the U.S. will take you to a site that says warranty registration isn’t open for U.S. customers. Instead, you have to go directly to www.zotacusa.com in order to register. Zotac does have U.S.-based tech support and RMA processing, though, along with toll-free tech support via phone from 9AM to 6PM Pacific.
EVGA’s take on the GTX 260 Reloaded has the same memory clocks at the Zotac, but its core and shader clocks are a little more conservative at 626/1350MHz. The EVGA costs a little more, too, at $309.99, but you do get the comfort of 24×7 toll-free tech support and EVGA’s Step-Up plan, which allows users to trade in their cards for credit toward an upgrade within 90 days of purchase. Like Zotac, EVGA offers a lifetime warranty, but only if you register within 30 days. Otherwise, the term is just one year. EVGA doesn’t bundle a game with its card, but folks who register via its website will get a free copy of 3DMark Vantage Advanced Edition, which, frankly, ain’t GRID.
Despite the GTX 260 Reloaded’s gains in GPU power, the cards themselves still require only dual 2×3-pin PCIe power plugs, unlike the GTX 280, which combines one 2×3 with one 2×4 connector.
In addition to the examples above, several other brands are selling GTX 260 Reloaded cards for around 300 bucksless in a few cases. For instance, Zotac’s much lower-clocked AMP! (not squared) Edition card is $289.99, although it’s hard to see the point unless you’re on a very specific budget. The new GTX 260s are pushing down prices on the older, eight-cluster versions, as well. MSI is selling an eight-cluster GTX 260 for $239.99, and it comes with a $40 mail-in rebate.
The 4870 doubles up on RAM
AMD’s answer to all of this is the card pictured below:
Pretty much looks like any old Radeon HD 4870, but the difference is simple: this one has twice as much GDDR5 memory onboard, a full gigabyte. We’ve been reviewing video cards with a gig of RAM for a while now and puzzling over when that much memory would really be necessary. Finally, we’re starting to see cases where 512MB clearly won’t suffice, and the Radeon HD 4870 GPU is powerful enough to perform well in some of those situations when given enough memory, as you’ll see in the following pages. Even if having more RAM onboard is largely future-proofing for upcoming games, AMD arguably needed to match the competition, and the GTX 260 has had 896MB of RAM since its introductionclose enough to a gig for most intents and purposes and well more than the 512MB on the first 4870s.
Although our example of the 4870 1GB is a reference card from AMD, a number of Radeon board vendors are now selling these cards. Like many others, Diamond’s rendition is $299.99 at Newegg, no rebate required. Trouble is, Diamond’s warranty term is only a year, and even that short window of coverage becomes invalid if you don’t register the product online within 30 days. So, yeah. That’s as much fun as impromptu dentistry. You might do better by going with the Asus version for $289.99. Asus covers its cards for three years andmiracle of miracles!tracks them via serial number, so no registration is required to get warranty service.
I really don’t care to be talking so much about warranty terms and mail-in rebates, but these things have become bigger issues in recent years. Mail-in rebates have spread like a cancer, especially among Nvidia’s partners, making video card pricing anything but straightforward. Meanwhile, board vendors have skimped on support, introducing these bogus warranty registration requirements. Both tactics rely on the same insight: inevitably, many customers won’t fulfill the exact paperwork requirements, thus saving the vendor some money. In some some cases, both rebates and warranties seem to require original receipts, which are more or less non-existent when you’re buying something online. So here we are, trying to help you tiptoe through a minefield. Props to Asus for doing the right thing here, and boo to Diamond for doing the exact opposite.
As one might expect, the 512MB versions of the 4870 are dropping in price as the 1GB cards arrive. Quite a few of them are available for $269.99, some with rebates attached on top of that, if you enjoy games of chance.
All of which sets us up for a renewed comparison of the Radeon HD 4870 1GB with the “reloaded” GeForce GTX 260. Both look like good options, and the real values may be in the cards they’ve essentially replaced and pushed down in price. As is our custom, we’ve included a whole raft of cards for comparison, ranging from the lowly GeForce 9500 GT to the exotic Radeon HD 4870 X2.
Our testing methods
As ever, we did our best to deliver clean benchmark numbers. Tests were run at least three times, and the results were averaged.
Our test systems were configured like so:
|Processor||Core 2 Extreme QX9650 3.0GHz|
|System bus||1333MHz (333MHz quad-pumped)|
|North bridge||X38 MCH|
|Chipset drivers||INF update 220.127.116.119
Matrix Storage Manager 7.8
|Memory size||2GB (4 DIMMs)|
DDR2 SDRAM at 800MHz
|CAS latency (CL)||4|
|RAS to CAS delay (tRCD)||4|
|RAS precharge (tRP)||4|
|Cycle time (tRAS)||12|
with RealTek 18.104.22.16818 drivers
4670 512MB GDDR3 PCIe
Diamond Radeon HD
3850 512MB PCIe
with Catalyst 8.8 drivers
| Asus Radeon HD 4850 512MB PCIe
| Diamond Radeon HD
4870 512MB PCIe
with Catalyst 8.9 drivers
| Radeon HD
4870 1GB PCIe
with Catalyst 8.9 drivers
| Palit Radeon HD
4870 X2 2GB PCIe
with Catalyst 8.9 drivers
Zotac GeForce 9500 GT ZONE
512MB GDDR3 PCIe
with ForceWare 177.92 drivers
with ForceWare 177.92 drivers
with ForceWare 177.92 drivers
with ForceWare 178.13 drivers
Zotac GeForce GTX 260 (216 SPs) AMP²! Edition 896MB PCIe
with ForceWare 178.13 drivers
with ForceWare 178.13 drivers
|Hard drive||WD Caviar SE16 320GB SATA|
|OS||Windows Vista Ultimate x64 Edition|
|OS updates||Service Pack 1, DirectX March 2008 update|
Thanks to Corsair for providing us with memory for our testing. Their quality, service, and support are easily superior to no-name DIMMs.
Our test systems were powered by PC Power & Cooling Silencer 750W power supply units. The Silencer 750W was a runaway Editor’s Choice winner in our epic 11-way power supply roundup, so it seemed like a fitting choice for our test rigs.
Unless otherwise specified, image quality settings for the graphics cards were left at the control panel defaults. Vertical refresh sync (vsync) was disabled for all tests.
We used the following versions of our test applications:
- Call of Duty 4: Modern Warfare 1.7
- Crysis Warhead
- Half-Life 2 Episode Two
- Enemy Territory: Quake Wars 1.5
- Race Driver GRID 1.2
- 3DMark Vantage 1.0.1
- FRAPS 2.9.4
- PowerDVD 8.0.1830.50
The tests and methods we employ are generally publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.
The theoryand practice
To see how the changes to the GeForce GTX 260 affect its position in the grand scheme, here’s a comparison of some recent graphics cards. Note that we’ve derived these numbers from the actual clock speeds of the cards we’re testing rather than from the stock clock speeds established by the GPU vendors.
GeForce 9500 GT
GeForce 9600 GSO
GeForce 9600 GT
GeForce 9800 GT
|GeForce 9800 GTX+||11.8||47.2||23.6||70.4||470||705|
|GeForce 9800 GX2||19.2||76.8||38.4||128.0||768||1152|
|GeForce GTX 260||16.1||36.9||18.4||111.9||477||715|
|GeForce GTX 260 216 SPs||18.1||46.7||23.3||117.9||607||910|
|GeForce GTX 280||19.3||48.2||24.1||141.7||622||933|
|Radeon HD 4650||4.8||19.2||9.6||16.0||384||–|
|Radeon HD 4670||6.0||24.0||12.0||32.0||480||–|
|Radeon HD 3850||11.6||11.6||11.6||57.6||464||–|
|Radeon HD 4850||10.0||25.0||12.5||63.6||1000||–|
|Radeon HD 4870||12.0||30.0||15.0||115.2||1200||–|
|Radeon HD 4870 X2||24.0||60.0||30.0||230.4||2400||–|
Like I said before, the GTX 260 Reloaded comes very, very close to the original GeForce GTX 280. By contrast, the Radeon HD 4870’s theoretical throughput is unchanged by the addition of more RAM.
In theory, it would seem that the GTX 260 Reloaded should have the edge in fill rate and texturing capacity, while the 4870 1GB ought to have more shader power. However, these things are often complicated in actual use by the quirks and varying efficiencies of the GPU architectures in question. Here’s how things shake out when we measure them with 3DMark’s directed benchmark tests.
This is pretty much the inverse of what we expected: the 4870 1GB proves faster in the color and texture fill rate tests, while the GTX 260 Reloaded takes three out of the four shader processing tests. Funny how that works. Another intriguing result: the GTX 260 Reloaded outruns the GTX 280 in the GPU cloth and particles benchmarks. This result suggests Nvidia may be using only one or some subset of the GT200’s thread processing clusters for certain types of work, such as vertex processing. In that case, the GTX 260 Reloaded’s higher shader clocks would matter more than the GTX 280’s additional TPC.
Call of Duty 4: Modern Warfare
We tested Call of Duty 4 by recording a custom demo of a multiplayer gaming session and playing it back using the game’s timedemo capability. We’ve chosen to test at display resolutions of 1280×1024, 1680×1050, and 1920×1200, which were the three most popular resolutions in our hardware survey. We enabled image quality enhancements like 4X antialiasing and 16X anisotropic filtering, and we also added 2560×1600 to the list for the very fastest cards, in order to really stress them.
Nvidia gains some ground on AMD here, since CoD4 doesn’t appear to benefit much at all from the 4870 1GB’s additional video memory. The GTX 260 Reloaded is true to form, performing quite a bit better than the original GTX 260and almost as well as the GTX 280. That’s enough of a boost to push the GTX 260 Reloaded past the 4870 1GB, although you probably would only notice the difference at 2560×1600.
Half-Life 2: Episode Two
We used a custom-recorded timedemo for this game, as well. We tested with most of Episode Two‘s in-game image quality options turned up, including HDR lighting. Reflections were set to “reflect world,” and motion blur was disabled.
The trend says the GTX 260 Reloaded scales a little better to the highest resolutions here, but we’re averaging over 60 FPS at 2560×1600, so I’m not sure it matters much.
Enemy Territory: Quake Wars
We tested this game with “high” settings for all of the game’s quality options except “Shader level” which was set to “Ultra.” Shadows and smooth foliage were enabled, but soft particles were disabled. Again, we used a custom timedemo recorded for use in this review.
Once more, even the highest resolution tested isn’t enough to confound the high-end cards. Still, the GTX 260 Reloaded slightly outperforms the Radeon HD 4870 1GB at higher resolutions.
Rather than use a timedemo, I tested Crysis Warhead by playing the game and using FRAPS to record frame rates. Because this way of doing things can introduce a lot of variation from one run to the next, I tested each card in five 60-second gameplay sessions. The benefit of testing in this way is that we get more info about exactly how the cards performed, including low frame rate numbers and frame-by-frame performance data. The frame-by-frame info for each card was taken from a single, hopefully representative play-testing session.
We first used Warhead‘s “Mainstream” quality level for testing, which is the second option on a ladder that has four steps. The “Gamer” and “Enthusiast” settings are both higher quality levels.
Both the GTX 260 Reloaded and the 4870 1GB perform acceptably here, with minimum frame rates of 35 and 29 FPS, respectively. The GTX 260 Reloaded is faster, though, as is the GeForce 9800 GTX+, surprisingly enough.
Switching to a higher resolution and quality level tends to clarify things a little, and what we learn is intriguing. The 4870 1GB appears to benefit from its additional video memory here. Also, the Radeon HD 4870 X2 really seems to struggle, just as it did in our first Warhead test. Apparently, AMD doesn’t have CrossFire working properly with this game yet.
Race Driver GRID
I tested this absolutely gorgeous-looking game with FRAPS, as well, and in order to keep things simple, I decided to capture frame rates over a single, longer session as I raced around the track. This approach has the advantage of letting me report second-by-second frame rates for our entire test session. All of the game’s quality settings were maxed out while we tested.
Check out that frame rate plot for the Radeon HD 4870. That’s a classic example of what happens when you run out of video memory. GRID was virtually unplayable at certain points around the track, where frame rates dropped deep into the single digits. The exact same Radeon HD 4870 GPU with 1GB of memory, however, produced performance easily superior to any of the GeForces.
Blu-ray HD video decoding and playback
One of the things that buying a new graphics card will get you that an older card or integrated graphics solution might not have is decode and playback acceleration for HD video, including Blu-ray discs. The latest GPUs include dedicated logic blocks that offload from the CPU much of the work of decoding the most common video compression formats. To test how well these cards perform that function, we used CyberLink’s new release 8 of PowerDVD, a Lite-ON BD-ROM drive, and the Blu-ray disc version of Live Free or Die Hard. Besides having the “I’m a Mac” guy in it, this movie is encoded in the AVC format (which includes H.264 video compression) at a 28Mbps bit rate.
In order to really stress these cards, we installed a lowly Core 2 Duo E4300 processor (a dual-core 1.8GHz CPU) in our test rig, and we asked the cards to scale up our 1080p movie to fit the native 2560×1600 resolution of our Dell 30″ monitor. We then recorded CPU utilization over a duration of 100 seconds while playing back chapter four of our movie.
The pattern we’ve seen before with low-end cards holds true here: the Radeons tend to be more effective at offloading the decode task from the CPU than the GeForces. The differences aren’t huge, though, and they’re accentuated by the slow CPU. With the usual fast quad-core processor in our test system, you’re looking at CPU utilization closer to 5% while playing a Blu-ray movie.
We measured total system power consumption at the wall socket using an Extech power analyzer model 380803. The monitor was plugged into a separate outlet, so its power draw was not part of our measurement. The cards were plugged into a motherboard on an open test bench.
The idle measurements were taken at the Windows Vista desktop with the Aero theme enabled. The cards were tested under load running Half-Life 2 Episode Two at 1680×1050 resolution, using the same settings we did for performance testing.
In spite of the fact that the GeForce GTX 260 Reloaded is based on a much larger chip, its power use is lower than the Radeon HD 4870’s, especially at idle, where Nvidia has worked wonders. The GTX 260 Reloaded-based system draws 30W less at the wall socket than our otherwise-identical system equipped with the 4870 1GB. AMD does seem to be making progress, though. The newer 1GB version of the 4870 draws less power at idle than the 512MB model.
We measured noise levels on our test systems, sitting on an open test bench, using an Extech model 407727 digital sound level meter. The meter was mounted on a tripod approximately 12″ from the test system at a height even with the top of the video card. We used the OSHA-standard weighting and speed for these measurements.
You can think of these noise level measurements much like our system power consumption tests, because the entire systems’ noise levels were measured. Of course, noise levels will vary greatly in the real world along with the acoustic properties of the PC enclosure used, whether the enclosure provides adequate cooling to avoid a card’s highest fan speeds, placement of the enclosure in the room, and a whole range of other variables. These results should give a reasonably good picture of comparative fan noise, though.
Some of the quieter cards, as you can see, fell below the ~40 dB threshold of our sound level meter, which robs us of exact data but points to a happy trend toward quiet coolers. The 4870 1GB and GTX 260 Reloaded were both able to register on our meter, but just barely. Neither card is annoyingly loud, and both exhaust hot air through openings in their expansion slot covers. Oddly enough, the GTX 260 Reloaded is quite a bit quieter than the GTX 280.
Per your requests, I’ve added GPU temperature readings to our results. I captured these using AMD’s Catalyst Control Center and Nvidia’s nTune Monitor, so we’re basically relying on the cards to report their temperatures properly. These temperatures were recorded while running the “rthdribl” demo in a window.
The 4870 1GB keeps alive the Radeon HD 4000-series tradition of high GPU temperatures. Its only close company on the Nvidia side of the aisle is a passively cooled GeForce 9500 GT and the GTX 280.
Despite the fact that these are tremendously complex chips with hundreds of millions of transistors, AMD and Nvidia have achieved a remarkable amount of parity in their GPUs. In terms of image quality, overall features, performance, and even price, the Radeon HD 4870 1GB and the GeForce GTX 260 “Reloaded” are practically interchangeable. That fact represents something of a comeback for Nvidia, since the older GTX 260 cost more than the 4870 and didn’t perform quite as well. If anything, the GTX 260 Reloaded was a smidgen faster than the 4870 1GB overall in our test suite.
The GTX 260 is based on a much larger chip with a wider path to memory, which almost certainly means it costs more to make than the 4870, but as a consumer, you’d never know it when using the two products, so I’m not sure it matters much for our purposes. Even the GTX 260’s power consumption is lower than the 4870’s, and its noise levels are comparable.
In the grand scheme, Nvidia may have a slight edge on AMD in some difficult-to-quantify ways. For instance, GeForce cards generally performed better with the newest game we tested, Crysis Warhead, and that’s not a surprising thing to see. Nvidia seems to do a better job of working with developers and ensuring good compatibility between its GPUs and new-from-the-box games. By contrast, the Radeons performed below our expectations in Warhead, and judging by the performance we saw from the Radeon HD 4870 X2, the CrossFire multi-GPU scheme isn’t yet working properly with this title. Nvidia may gain an additional advantage if and when we see PhysX-enabled games come to pass, but I wouldn’t factor that into a purchasing decision today.
On the other hand, AMD releases new drivers more often and has much better support for multiple monitors with CrossFire than the kludgy arrangement Nvidia uses with SLI. Also, CrossFire is broadly compatible with Intel chipsets, while Nvidia generally restricts SLI to nForce-based motherboards. So I dunno.
One thing I do know is that, whichever one you prefer, both of these cards are wicked fast for 300 bucks. In fact, you probably don’t need either one of them unless you plan on using it with a nice, big monitor with a resolution of at least 1920×1080 or 1920×1200. Heck, even then, you can get by very well in almost all of today’s games with something like a $170 Radeon HD 4850. What you get when you step up to one of these $300 cards is substantially more GPU power, memory bandwidth, and longevity potential. That may not always be apparent, but it may sometimes be painfully so. Only in a couple of casesWarhead at 1920×1200 and GRID at 2560×1600did we see the 4870 1GB’s extra memory make a difference versus the 4870 512MB. But the difference in GRID was night and day. Personally, given the choice, I’d pony up the extra cash for the 1GB card, just for the peace of mind. But I’m probably crazy for saying so.