Summary: 1600 stream processors. DirectX 11. Over 2 billion transistors. 3-Display support with Eyefinity. Those are just a few of the key features supported by ATI's latest Radeon 5800 series cards. Join us as we go over the tech behind this 3D behemoth and its performance in 10 of the latest games!
All warfare is based on deception. Hence, when able to attack, we must seem unable; when using our forces, we must seem inactive; when we are near, we must make the enemy believe we are far away; when far away, we must make him believe we are near. Hold out baits to entice the enemy. Feign disorder, and crush him.
- Sun Tzu, The Art of War
If your enemy is secure at all points, be prepared for him. If he is in superior strength, evade him. If your opponent is temperamental, seek to irritate him. Pretend to be weak, that he may grow arrogant. If he is taking his ease, give him no rest. If his forces are united, separate them. If sovereign and subject are in accord, put division between them. Attack him where he is unprepared, appear where you are not expected.
- Sun Tzu, The Art of War
These quotes, lifted from The Art of War -- ancient Chinese text written by Chinese general Sun Tzu thousands of years ago yet still required reading for military theorists and some business schools -- aptly describe ATI’s cunning gambit with Radeon HD 4000 series just over a year ago. Just when everyone was counting ATI out of the high-end graphics space, NVIDIA included, they delivered a homerun product with the Radeon 4850 and Radeon 4870 that stunned the world. You could make an argument that the ramifications were perhaps as significant as ATI’s first DirectX 9 product, R300 (Radeon 9700) was over seven years ago.
Think about it. As any hardware enthusiast who’s followed the industry can tell you, before the Radeon 4800 series cards arrived on the scene, graphics card prices were going nowhere but up. While it may seem hard to believe now, NVIDIA’s GeForce GTX 280 launched just two weeks before the Radeon 4870 at a staggering price tag of $649 MSRP. Its less powerful sibling, the GeForce GTX 260 carried an MSRP of $399.
That’s a steep climb up from the days when 3dfx’s original Voodoo Graphics card sold for $300. NVIDIA’s prior products, the GeForce 8800 Ultra and GeForce 8800 GTX sold for $829 and $599 respectively on their launch day.
Now before you think we’re out to get NVIDIA, they did give the mainstream segment one gem of a product in the form of the GeForce 8800 GT, and we said so when it debuted in back in October of 2007: just in time for Crysis. What ATI pulled off with the Radeon 4800 series though was even more significant than the debut of the 8800 GT.
If you rewind back to the days in early 2008 ahead of the Radeon 4800 and GeForce GTX 200’s debut, you’ll remember that all the early rumors indicated that the Radeon 4850 would fall somewhere between the 8800 GT and 9800 GTX in performance, with pricing similar to the 8800 GT, while the Radeon 4870 would outperform 9800 GTX. If you’re NVIDIA and you know you’ve got a performance behemoth like GT200 right around the corner, you’d be feeling pretty good about yourself and how your upcoming product will fare against your competitors. Some ATI fanboys were already blaming AMD for the supposed “death” of ATI’s high-end graphics and were ready to queue up Taps.
Ultimately what ATI ended up delivering with the Radeon 4850 and 4870 was completely different in a good way. Priced at $200 (the same price NVIDIA’s 8800 GT sold for at the time), the Radeon 4850 delivered performance that was greater than the 9800 GTX, forcing NVIDIA to concoct the 9800 GTX+ as a counter to the 4850. Meanwhile, the Radeon 4870 had its sights set on the GTX 260, yet it was priced $100 less. NVIDIA was forced to counter this GPU with GTX 260 price cuts and rebate checks for early adopters who picked up GTX 260 and 280 cards ahead of the arrival of the new Radeons.
To this day the 216-shader GeForce GTX 260 and Radeon 4870 1GB are still cutthroat competitors.
ATI’s “sweet spot” strategy of delivering smaller, more cost effective gaming GPUs for the performance segment and then scaling that tech up and down for the high-end and value markets was a real game changer for the industry, and their execution on this strategy was executed as if it was masterminded by Sun Tzu himself.
Enough about the past though. Now its time for the dawn of a new era of DirectX 11 gaming. Given the success of their Radeon 4800 series, ATI wouldn’t have the benefit of surprise this time around. Instead they’re essentially using lessons learned with GDDR5 memory, the development of DirectX 10.1 hardware, and TSMC’s 40-nm manufacturing process to give them a time to market advantage over the competition,
Being first to market didn’t help Sega’s Dreamcast, but it’s worked wonders for Microsoft’s Xbox 360 (I know, it’s a terrible analogy that oversimplifies things, but it’s the first gaming-related comparison I could come up with). Which scenario will ATI’s next-generation Radeon 5870 ultimately end up? We can’t answer that question today, but we can tell you how it looks and performs with today’s games…not to mention its breakthroughs in noise and power consumption…and don’t forget the OC’ing. Read on for the full details!
If the RV770 GPU inside the Radeon 4850 and Radeon 4870 was ATI’s first TeraScale graphics engine (an obvious nod to the GPU’s distinction as the first desktop graphics card to break the 1 TeraFLOP mark), then RV870’s TeraScale 2 has to be twice as good right? Right. In fact, Radeon 5870 more than doubles the compute power of Radeon 4870 (1.2 TeraFLOPS), boasting up to 2.72 TeraFLOPS:
Radeon 5870 Specifications
TeraScale 2 Unified Processing Architecture
GDDR5 interface with 153.6 GB/sec of memory bandwidth
PCI Express 2.1 x16 bus interface
DirectX 11 support
OpenGL 3.2 support
Image quality enhancement technology
ATI Eyefinity multi-display technology
ATI Stream acceleration technology
ATI CrossFireX™ multi-GPU technology
ATI Avivo HD Video & Display technology
Integrated HD audio controller
Speeds and feeds
2.15 billion transistors w/334mm2 die size
40-nm manufacturing process
256-bit GDDR5 memory interface
500 Watt or greater power supply with two 6-pin PCI Express® power connectors
recommended (600 Watt and four 6-pin connectors for ATI CrossFireX™ technology in
World’s first DX11 GPU
Besides the Radeon 5870’s 1600 stream processing units, a lot has been made about DirectX 11. Honestly DirectX 11 is more of an evolutionary extension of the concepts introduced in DirectX 10 more than anything else. It isn’t the ground up rewrite that DX10 was.
But it does bring with it a number of important new features such as new shader model 5.0 instructions designed to deliver simplified coding, better edge detection for anti-aliasing, and faster shadow filtering and ambient occlusion. DirectX 11 also offers support for multithreading: the DirectX driver, runtime, and game can all run in their own separate threads. But the most talked about additions are probably the new tessellation unit and compute shader.
ATI’s offered an integrated tessellation unit since the R600 GPU powering the Radeon 2900 XT, but the DX11 tesellation unit is more flexible than ATI’s. The compute shader is Microsoft’s answer to GPGPU solutions like OpenCL, CUDA, and ATI’s Stream computing initiative. The implications for gamers could be huge; the compute shader could potentially be used by game developers to bring GPU-based physics, ray-tracing, better AI, and more.
Microsoft has a ton of information on DirectX 11 which can be found here if you’d like to read up more on the subject. You can also check out ATI’s slides above. Honestly we would’ve liked to have had a dedicated DirectX 11 article up just prior to the arrival of DX11 hardware (similar to what we did with DirectX 10 a few years back), but obviously new CPUs from AMD and Intel launched earlier this month prevented that from happening.
If there’s only one key word to take away from the architecture of the new RV870 chip found inside ATI’s Radeon 5800 series cards it’s “2X”. Thanks to the smaller 40-nm manufacturing process, ATI can afford to double up on pretty much everything that made RV770 so special a year ago without having to charge double the price.
As you just saw on the specs page, RV870 boasts twice the SIMD units as its predecessor, RV770. Each SIMD unit consists of 80 stream processors and one texture unit, so with twice the SIMD cores, you’ve got twice the number of stream processors (1600) and twice the number of texture units, 20 (80 effective). Now ATI and NVIDIA use different nomenclature for what they both describe as a “stream processor” -- the actual physical number of stream processing units inside RV870 is actually 320 -- but regardless of the term you use to describe them, it’s an impressive amount of compute power nonetheless, as the 5870’s 2.72 TeraFLOPS can attest to.
The overall layout of RV870 is similar to RV770, just bigger. See for yourself in this RV870 block diagram:
The 20 SIMD cores are depicted as the red squares in the center of the diagram. If you look a little closer, you can sit and count the individual stream processing units for yourself. Like RV770 each stream processing unit consists of 4 stream cores+1 special function stream core which are tied to a branch unit and general purpose registers. ATI has tweaked them to improve their IPC.
Tied to each SIMD core is its own dedicated texture unit, again, just like RV770. RV870 boasts improved texel fill rate, up to 68 (bilinear filtered) Gigatexels/sec and improved data fetch rate: up to 272 billion fetches/sec. ATI’s also improved the cache bandwidth of the L1 texture caches tied to the texture units. RV870 sports up to 1 TB/sec L1 texture fetch bandwidth, while peak bandwidth between the L1 and L2 caches tops out at up to 435GB/sec.
In comparison, RV770 featured up to 480GB/sec of L1 texture fetch bandwidth and up to 384GB/sec of bandwidth between the L1 and L2 caches.
Up at the top of the block diagram, you’ll notice another significant tweak ATI has made with RV870 is the addition of a second rasterizer in the graphics engine of the chip. With a second rasterizer, RV870 feeds more pixels into the engine than its predecessor; this is important when you’re dealing with a GPU that’s outfitted with 1600 stream processors. ATI’s also updated their tessellator for DirectX 11 compliance.
256-bit memory interface
Moving to the bottom of the block diagram, you’ll also see RV870’s four 64-bit memory controllers, just like RV770. This probably comes as a bit of a disappointment to those of you who were hoping for a wider memory interface and the potential performance boost it could bring under high resolution, high AA scenarios (especially since some of the rumor sites were saying earlier this summer that RV870 would possess a wider memory interface), but in speaking with ATI, they felt that a 256-bit interface with high-speed GDDR5 was the way to go given their die size and transistor budget constraints.
Besides is smaller manufacturing process, which naturally helps to reduce the GPU’s power consumption, ATI’s also integrated tweaks to further reduce RV870’s power consumption. As listed on the specs on the previous page, the chip consumes as little as 27W at idle. Impressive for a GPU to contains over 2 billion transistors.
Like the RV770 launch last year, ATI’s providing two SKUs based on their RV870 chip: the Radeon HD 5870 and Radeon HD 5850. Both SKUs utilize the exact same RV870 chip, only in the case of the Radeon 5850 two SIMD cores are disabled. This drops the 5850 down to just 1440 stream processors and 72 (effective) texture units. Clock speeds are also reduced as well, although the 5850 retains the use of GDDR5 memory. If you recall, none of the SIMD cores were disabled on the Radeon 4850, but ATI utilized slower clocks and slower GDDR3 memory
The following chart compares the two SKUs to each other, as well as their direct predecessor, Radeon 4870:
As you can see in the chart above, the Radeon 5850 will sell for approximately $259. That’s $60 more than its direct predecessor, the Radeon 4850. This has upset some ATI enthusiasts who were expecting ATI to hold the line at $199 and $299 for this generation also. The key difference here however is that the Radeon HD 5850 will deliver performance that’s greater than ATI’s previous flagship, the Radeon 4890, which currently sells for around $200 (plus or minus $10 depending on card configuration and rebates). In addition, remember that the 4890 launched at $249 earlier this year, so ATI’s only charging a premium of $10 over that price.
Considering the Radeon 5850’s superior performance, ATI’s decided to slot it above the Radeon 4890 in the price stack.
Priced at $379, the Radeon 5870 is ATI’s new flagship product, and while it’s obviously priced higher than the Radeon 4870 was on its launch day, it’s also cheaper than the 4870 X2 was when it debuted and should deliver better performance in some situations, with the obvious addition of DirectX 11 and no longer having to worry about whether or not the latest game you want to play has a proper CrossFire profile or not.
While both GPUs are officially launching today, you’ll only be able to purchase the Radeon 5870 right now. Unfortunately, this isn’t a “hard launch” for both GPUs. In addition, supplies are going to be tight initially. ATI says they’ve worked with TSMC to resolve the issues that plagued their previous 40-nm product, Radeon 4770, but obviously this launch isn’t going to be as smooth as the Radeon 4890 and Radeon 4870/4850 introductions were with regards to supplying enough cards to meet demand.
ATI made a huge splash with their Eyefinity announcement earlier this month, which brings support for up to three displays to the Radeon 5870 and Radeon 5850. ATI says that a special Radeon 5870 Eyefinity Edition will be coming later this year with support for up to six displays. Presumably this is the 2GB Radeon 5870 SKU rumor sites have been writing about, as today’s Radeon 5870 card’s are all outfitted with 1GB of memory. The 5870 Eyefinity Edition will ship with six mini-DisplayPort outputs on its backplate.
For the full 3-display Eyefinity experience, at least one display must utilize DisplayPort. Two HDMI, VGA, or DVI-based displays can then be combined with the DisplayPort display in any combo you’d like (two DVIs, one DVI/one HDMI, etc).
If you don’t have a DisplayPort display but you do have three DVIs, you’ll have to purchase an active DisplayPort-to-DVI adapter with external power. ATI says an active adapter is needed to “convert the DisplayPort signal from the graphics card, modify it to the new display signal required for the attached monitor, and for transmission.” Currently there’s only one such adapter on the market, Apple’s $109 adapter, but ATI says they’re working with more manufacturers to bring less expensive solutions to market beginning in Q4 of this year.
We hope to provide a dedicated article on Eyefinity and its gaming ramifications when we have more time in the coming weeks.
Radeon 5000 series roadmap
Besides the launch of today’s RV870 “Cypress” chip, ATI says they have two more GPUs that will be debuting in Q4 of this year. For the high-end enthusiast who wants to fastest card money can buy, AMD will be offering “Hemlock”.
If you recall the original Radeon 4870 and Radeon 4850 reference board designs, you’ll remember that the stock 4850 relied on a single-slot heatsink/fan unit while the 4870’s dual-slot cooler reminded us of the implementation used on the 2900 XT at first glance.
ATI didn’t repeat the same mistake with follow-up GPUs like the Radeon HD 4890 and 4770. Both of these cards shipped with quite component coolers in stock form.
The cooling solution ATI has come up with for the Radeon 5870 takes this to another level though.
As you can see, ATI has come up with a new dual-slot ducted design for the 5800 series cards. The new duct isn’t completely enclosed, you’ll find vents around the back of the card, just behind the fan, as well as additional cooling vents near the CrossFire connectors. An exhaust vent is then located on the back plate of the card. In order to make room for all the display outputs on the back of the card, this vent is a little smaller than previous designs.
Because of this new design though, some air from your GPU will vent into your system case. The majority of the air does seem to exhaust out the back of the card though.
To up the performance factor of this new cooler, ATI has added a fourth heatpipe to the 5870’s cooler. In comparison, the 4870 was equipped with just two heatpipes, while the 4890 sports three heatpipes. Like previous coolers, the heatpipes are made from copper to increase their effectiveness at drawing heat off the GPU. A large, dual-slot aluminum heatsink then helps to keep everything cool.
While the fan itself looks like a carryover from the 4890, it’s actually been outfitted with quieter bearings than the old cooler. To further reduce noise levels, the fan itself spins at lower RPMs than the 4890 did: 1200 RPMs at idle versus 1600. As a result, the card is literally whisper quiet at idle, and even under load we couldn’t get a reading from our Extech sound level meter.
Short of an all-passive card, this is definitely the quietest reference board we’ve seen from any manufacturer. ATI’s partners will be hard-pressed to match it when it comes to performance versus noise output. We observed idle temps as low as 50 degrees Celsius with the 5870, while load temps topped out at 68 degrees. Not bad at all for a reference design.
Those of you with smaller cases though won’t have a fun time housing this card inside your system. Measuring in just shy of 11 inches, the Radeon 5870 is an incredibly long card. That makes it longer than the GTX 200 series and on par with dual GPU boards like the Radeon 4870 X2 and GeForce GTX 295.
While we haven’t received a Radeon 5850 board yet, we’re told that it measures around 9.5”, the same as today’s Radeon 4890.
Fortunately the power requirements aren’t insane. The board needs just two 6-pin PCIe power connectors, with a 500W power supply the minimum recommendation.
Moving around to the backplate of the card, you’ll find two dual-link DVIs, as well as DisplayPort and HDMI outputs.
Like previous high-end GPU releases from ATI, partner boards will initially be limited to ATI’s reference board design only. As such, the hardware itself will be the same amongst the various card manufacturers. Clock speeds will also be limited to the standard reference specifications from ATI. In other words, you won’t see factory OC’ed 5870 cards from any of ATI’s launch partners today.
XFX for instance will continue to provide their double lifetime warranty for all of their Radeon 5870 and 5850 cards. This popular feature provides lifetime warranty coverage for the original card owner, as well as the second card owner (say for instance the original owner decides to throw his card up for sale on eBay a few years from now) as long as they’re both registered with XFX.
ASUS will offer their boards with their exclusive Voltage Tweak technology, which allows you to increase the GPU/memory voltage to further aid you in your OC’ing endeavors.
Meanwhile, Sapphire hopes to have their first Vapor-X board ready sometime next month.
DiRT 2 included
If you recall the Radeon 9800 XT/9600 XT days, one incentive ATI included with all the cards of that era was a voucher to download Half-Life 2 once it was released. ATI is doing this again for the Radeon 5850 and 5870. Purchase a 5850 or 5870 board and you’ll get a voucher for the game so you can play it when it’s released this November.
3D Performance Testbed
Intel Core i7-975 Extreme Edition @3.33GHz
Power Consumption Testbed
Intel Core i5-750
Call of Duty: World at War
Fallout 3 DirectX 9
CoD: WaW – DirectX 9
Crysis – DirectX 10
Crysis – DirectX 10
Far Cry 2 – DirectX 10
STALKER Clear Sky – DirectX 10
Left 4 Dead – DirectX9
CoD: WaW – Direct3D9
Batman:AA – DirectX 9
So why are we testing a $379 graphics card without AA in Batman? It turns out that the game features an in-game AA option for GeForce cards, but requires ATI Radeon cards to use the control panel to enable AA. This puts ATI cards at an artificial disadvantage, as forcing AA via the control panel applies AA on everything within the scene, whereas in-game AA selectively applies AA on areas where the developer sees fit. ATI says the following about the issue:
Batman:AA – DirectX 9
HAWX – DirectX 10
Wolfenstein – OpenGL
Left 4 Dead
Resident Evil 5
Call of Duty: World at War
Unfortunately, the highest speeds Overdrive supports for the 5870 at this time are 900MHz core/1300MHz memory. These speeds were easily attained by both our 5870 boards. ATI says they’re looking into perhaps providing unlimited caps in a future driver release, but when, or even if that will happen is unknown. Enthusiasts will have to wait for apps like RivaTuner to add support for the RV870 GPU.
In our benchmarks today the Radeon 5870 card clearly delivered the best performance of any single GPU we tested. While performance wasn’t double that of Radeon 4890, the 5870 did post some impressive gains in the most graphically intensive games we test with: STALKER Clear Sky, Crysis, and Far Cry 2; generally running around 1.5 times faster than ATI’s previous flagship offering in these titles. In comparison to NVIDIA’s GeForce GTX 285, the 5870 generally ran 20-30% in the aforementioned games, although the 5870 really pulls away from the GTX 285 once Crysis is run under the “Very High” graphics setting. Here the Radeon 4890 is actually able to keep pace with NVIDIA, while the 5870 blows them both away.
In some of the other games the gains aren’t quite as impressive. In Call of Duty: World at War – based on an engine which has traditionally favored NVIDIA’s architecture in previous benchmarks – the GTX 285 trails the 5870 by just 11% at 2560x1600 resolution (the Radeon 4890 finishes behind the 5870 by 28% under this scenario). In another recent release, Capcom’s Resident Evil 5, the GTX 285 trails the Radeon 5870 by just 15% at 1600x1200 and 1920x1200. ATI says that they were “unable to receive builds of this game early enough to get a chance to test and address any open issues. We will work with the developer to test and adjust any compatibility or performance issues that we encounter.” That could explain part of the reason why the GTX 285 was able to hang so close to the 5870 at those lower resolutions.
Fans of NVIDIA will probably bring up our results with Wolfenstein too, but we’re not going to harp on these results as ATI’s OpenGL performance isn’t as big of an issue today as it was 5 years ago when people were actually playing OpenGL titles more frequently. Hop on a Wolfenstein server any night of the week and you’ll see just how dead the scene is today.
Overall ATI’s Radeon 5870 crushes today’s latest DX10 GPUs in performance, but you’re going to need to really crank up the settings and run newer, more demanding titles to see where the card really shines.
It’s the multi-GPU cards like the Radeon 4870 X2 and GeForce GTX 295 as well as the SLI/CrossFire setups that are going to be the biggest threat to the Radeon 5870 out-of-the-gate. While ATI’s Radeon 5870 is clearly the fastest GPU on the planet right now, NVIDIA’s GeForce GTX 295 is still the world’s fastest graphics card overall. The 5870 manages to outrun the GeForce GTX 295 in H.A.W.X., but like Wolfenstein, this is one of those unpopular games that’s received a lukewarm reception from the public so far. ATI leverages DirectX 10.1 to outperform NVIDIA’s offerings, just as NVIDIA’s more efficient OpenGL driver allows them to outrun the ATI cards (with the obvious exception being the 5870) in Wolfenstein.
Because the 5870 isn’t delivering 2X performance speedups over the prior generation at this point, those of you who already own Radeon 4870 1GB or Radeon 4890 cards may want to just pick up a second card for CrossFire while they’re still cheap and plentiful. Radeon 4800 prices really can’t go any lower, and reportedly these cards will be phased out to make room for Juniper cards which will come in the next few months.
This story’s not anywhere near being over though. We’re pretty convinced that ATI’s driver team still has quite a bit of performance to wring from the hardware. Radeon X1800 owners remember the remarkable speedups follow-up Catalyst drivers were able to deliver, and ATI pulled off the same trick for Radeon 4800 series card owners last year. It’s definitely conceivable that in another 3-6 months the 5870 could end up delivering the 2X performance speedups you expect from a next-generation product.
ATI’s Radeon 5850 is another topic we can’t wait to write about, and don’t forget about the first DX11 games coming later this year like DiRT 2 and STALKER: Call of Pripyat, as well as ATI’s innovative Eyefinity technology.
Yes ATI, the game has changed, and you’re now the technology and performance leader. Clearly NVIDIA’s probably encouraged by the early performance results for 5870 though. Their GeForce GTX 295 is still the world’s fastest card, and while it’s now based on outdated technology, until the first DX11 titles show up, DirectX 11 is only an advantage that exists on paper.
|© Copyright 2003 FS Media, Inc.|