Radeon HD 6850 and 6870 Review

Introduction

SPECIAL NOTE:

The following article will reference pictures that are not there. This is not an accidental omission on my part. I am in the process of buying a new camera, but I wanted to start getting reviews posted and get the site going without having to wait. Please be patient, and check back often as I will post when I get the camera, and let you know that the images are going to start coming. Thank you for your patience.

Note: The 6950 and 6970 are forthcoming by the end of next week. I will update this review at that point, with fresh benches and commentary.

 

Ad Note: For this article, I am experimenting with Google AdSense. I am looking for feedback from readers as to whether you are willing to read articles with AdSense ads, and how willing you are to click on them, either to augment a premium subscription, or in lieu of it. If each reader commits to clicking on one or two ads per article each visit, I could conceviably do the site without the premium content. For now please send me a message with your thoughts.

Overview of the 6x00 Cards

First impressions

At first glance, these cards look identical in size and overall feel to the 5x00's. The casing for the cooling is all new, and whether it's an improvement is a matter of taste. The temporary camera I have available isn't suited to close up shots, so I'll spare you the blurries right now. In a few weeks, I will have my new DSLR, and I will update most of my reviews with fresh, high quality, macro images.

The Architecture

The New Radeon Generation

The first thing that most people ask, and was certainly the first thing I wanted to know was: What's with this confusing naming scheme?

In terms of the progression of naming schemes in things such as PC components and even cars, we expect a certain numerical progression to denote a lineage of product line. Therefore, this is what we expect:

Radeon 5850 begets Radeon 6850

Radeon 5870 begets Radeon 6870

Radeon 5970 begets Radeon 6970

And then the world would keep on turning.

This isn't the case with the new generation of Radeons. So why the change in naming scheme? Let's take a look at the 5x00 cards, and we'll see how the performance scale generally works out:

From Entry level to high end, we have the following, in ascending order:

• 5450
• 5550
• 5570
• 5670
• 5770
• 5830
• 5850
• 5870
• 5970

That's quite a few SKU's. Right away we notice a few things. First, there are 9 levels of card, and with OEM variations in clocks and features, that's a bewildering array of choices. I have a feeling that AMD looked at this as an opportunity to make the 6x00 series a lot more intuitive to judge and if that made the transition a little funky, that's a reasonable tradeoff.

The Radeon 5000 Generation

For modern, DirectX 11 games, anything below the 5570 really isn't enough card to produce playable framerates. The 5570 is a $70 price point, and it's not going to blow through games on high settings or even low settings with a lot happening. So that leaves the next level up, which to me is really the 5850. The intermediate cards aren't really a good value over the 5570 to justify their existence and really only serve to complicate an already complicated buying process.

The 5850 can currently be had at about $200, and represents what enthusiasts would call the mid-range; enough to play most games on high settings at smooth frame rates.

The 5770 produces good results, and is priced between the two, but I feel that if you are going to spend around $150, you should spring for the 5850, which will do much better at higher resolutions and detail settings, and represents a more forward looking investment.

The 5870 is the fastest single chip solution of this series, and at around $250 is a significant investment. To this day it will play any modern game on full details at HD resolutions, and will continue to do so for some time to come. Even when it was $400 it was the best game in town for a Radeon single chip and worth every penny.

The king of the road, to this day, is the Radeon 5970, essentially two 5870's on one PCB. As we'll see below, this card still destroys everything else and commands a price to match. Still selling as high as $700, but if you need the best, this is it.

So where does this leave us? With a confusing and massive array of cards to choose from. How about we just simplify?

The Radeon 6000 Generation

I imagine that AMD, in the middle of a branding transition from ATI to pure AMD, had to look at all of this and think: We have to accomplish three goals:

Simply the product line

Better differentiate the performance goals of each level of product, shedding away stuff like the 5830

Reinvent the product image to better match the new AMD only branding

How did they do?

Many early reviews were flummoxed by the new confusing name scheme, but as is usually the case with my reviews, I like to take a step back and look at the big picture, the HOW's and WHY's of things. I do not have any inside information. I have no internal memos, or interviews to go on, just my instincts.

The above examples represent what I think are the key goals that would be served by changing the expectation of names for the products. How do I think they did? Brilliantly. So long as AMD resists the urge to introduce 12 new variants of these same cards, I think we're going to forget about the alteration and just move on with comparing these products to each other, to Nvidia, and to the inevitable 7000 series.

New Features

The 6000 series Radeons introduce some new features to the mix. How important they are depends on your particular usage, but since they do not add to the cost of the product, they are value added.

3D Video Support

UVD 3.0

HDMI 1.4a

More standard ports for Eyefinity

We'll look at each of these features in more detail in a follow on review, once I have the 69x0's in hand. For now, let's take a look at the new cards, the new name scheme, and how it all fits together.

Price or Name: Which one do I pick now?

Earlier I said that the new cards do not beget the same general numbering scheme. This is true if you look at it from a purely number perspective, so we have to look at things from a newer, more accurate price/performance/features perspective. We'll see in the end that it all makes sense; we just have to adjust our expectations to include the big picture, rather than just the frames per second, which is only part of the job of a GPU.

How Price breaks down: ( from the perspective of playing DirectX 11 games )

Radeon 5570 - $70

Radeon 6850 - $200

Radeon 5850- $200

Radeon 6870- $250

Radeon 5870- $300

Radeon 5970- $700

- - -  The following cards will be in my lab next week, and I will review them at that point

Radeon 6950- $300

Radeon 6970- $370

Looking at the above, it already becomes clear that the 6x00 cards have a very clear division of price between each class, or roughly $50. Why no lower end cards? We really don't need any right now, several of the 5x00 cards fill the under $200 slots nicely, as well as filling in price gaps in the other ranges. It's elegant. As for performance, we'll get into that in a moment, but once we do, you will see that from the bigger picture "price/performance/features" viewpoint, the 6x00 cards just destroy the old 5x00's as it should be. Many early reviewers jumped to the conclusion that AMD doesn't know what they are doing, and may have been hiding the fact that the 6x00's do not perform up to spec, but this is simply not true. We must change our expectation and get with the program here.

Test Platform and Methodology

1156 Test Platform

CPU - Intel Core i5 661 overclocked to 4.2Ghz. It's rock stable, so I didn't bother to do stock settings for this round.

CPU Cooling - Coolit System E.C.O. 120

Motherboard - Gigabyte GA-P55A-UD3

RAM - G.Skill Trident F3-16000CL9T-6GBTD x2 ( 4GB total )

PSU - OCZ ModXStream Pro 700w

OS - Windows 7 x64

All told this represents a fairly midrange system by current standards depending on GPU used. I would skip using the 5970 for this platform because I feel that the dual GPU card is overkill for this build and does not represent a fair assessment of performance. The gigabyte board did not handle the 5970 very well, but I am including the numbers for it just to be a completest.

1366 Test Platform

CPU - Intel Core i7 970 at stock clocks. I also have 4.2+Ghz results, but I will do another set of results for that later.

CPU Cooling - Coolit System E.C.O. 240 with Push/Pull Fans

Motherboard - Asus Rampage III Extreme

RAM - OCZ Reaper OCZ3RPR1333C9LV12GK x3 ( 12GB total )

PSU - Corsair AX1200 1200w

OS - Windows 7 x64

No doubt about it, this is a HIGH end system. Full reviews of key components are forthcoming, but for now, what you need to know is that this system is not a bottleneck for any card.

Test Results

Gigabyte Motherboard with Core i5 661 running 4.2Ghz

3Dmark 11 Extreme Setting

3Dmark 11 is the new standard for DirectX synthetic benchmarks. As such, I am going to open this inaugural bench test with this version, and am not going to use Vantage at all moving forward, so as to have a clear emphasis on DirectX 11, and to cut down on redundant data. Vantage and 11 results between cards are the same overall difference, so why not use the newest?

For Extreme settings, Futuremark settled on a resolution of 1920x1080, which is in line with HDTV standards and all newer PC monitors. The standard has been a 16:10 ratio, which is better for desktops, but when it came down to lowering manufacturing costs, it just made more sense for monitor makers to fall in line with TV standards. It's nice to see 11 following along with this trend as well. If you have a 1920x1080 monitor, Vantage will refuse to run the Extreme setting anyway, so it all works out.

With the Extreme setting, we see a very expected lineage of performance. At the bottom with a retail price under $70, the 5570 looks like it's getting stomped in this test, and it is, but as we'll see in the below tests, this card still provides very playable results in real world games.

The 6850, 5850, and 6870 show very little difference on this CPU. As we'll see below, the gap widens when you have a much more powerful CPU at work, and I purposely did this test with the dual core at 4.2Ghz, and the hexacore at 3.2Ghz to really show off this disparity.

The 5870 is a nice balance of power here, and finally I did the 5970 at stock clocks even though you'd never do a build with this combination, just to show that if you throw enough brute rendering muscle at a task you can make it work, even if you have to increase your cooling and power significantly to support it.

3Dmark 11 Performance Setting

Interestingly, once we dial it down to a more realistic resolution and performance expectation for the dual core platform, in this case the "Performance" setting for 3Dmark 11, which is 1280x720, with lower details, the card classes somewhat even out, and we see that the 6850 and 5850 perform about the same and the 6870 and 6850 perform about the same. Why is this interesting?

Price - The two card sets share a similar price class. The 6850 is currently selling for about $200 online, and the 5850 is also selling at $200. How can this be? It all boils down to the difference of expectation. Yes, the 5800's are faster than their 6800 counterparts ( at least in name scheme ), but the 6X00's have additional features not seen on the older cards such as 3D video support, which has long been a selling point of Nvidia cards.

The 6X00's have more standard ports, HD3D support, updated video acceleration, and UVD 3.0. For all around value, it's no contest, but if you're looking for more speed to dollar ratio it gets more complicated.

Batman Arkham Asylum Benchmark Average FPS

Arkham uses the ever popular Unreal Engine III ( as will Arkham City ), and although the Radeon cards do not support PhysX directly, the game renders quickly and beautifully on each card, which again shows the same expected curve as we move up the card classes. The real world difference between the middle four cards is nill. Unless you need PhysX, any modern Radeon card will play this game just fine.

Just Cause 2 – Dark Tower

Just Cause 2 uses a different engine from the others in this test, the "Avalanche Engine 2.0", which supports a massively sandbox world, robust physics, and a limited, but respectable level of distance rendering. On the dual core, we see the first anomalous results. The "Dark Tower" benchmark is a flythrough of a beach and tower at night, and shows off the beautiful water and particle effects of this engine. On the dual core, there is much less difference between the high end boards than we see on the hexacore, and other engines. We can attribute this to the CPU bottlenecking the effects, but as we see, the game is still quite playable even on the lowest end card.

Just Cause 2 – Desert Sunrise

The "Desert Sunrise" shows off a huge amount of effects, from shimmering light, sunrays, water reflections, heat waves, to weather. Here we see some odd results, with the 5870 pulling out ahead of the 5970. The difference is quite mild, and not as severe as the chart makes it look. Several tests bear this out, but it could be the result of a number of factors, but I'm going with power throttling through the board. Take a look at the 5970 results on the Rampage board and you'll see it back on top where it belongs.

Just Cause 2 – Concrete Jungle

"Concrete Jungle" is a flythrough of a massive cityscape, complete with huge reflections from glass buildings, progressive weather effects, and lots of water. A true "bring the GPU to its knees" test. In point of fact, this is the first time that the 5570 really gives up the ghost, and cannot produce close to 30 fps on this platform. Even on the Rampage it provides no more fps, as this test is truly more about GPU potency. Again, we see the 5970 being reined in needlessly. The Gigabyte board just isn't made for this kind of performance, being a midrange board. Still, you'd think it would be able to handle a 5970 at stock clocks without having issues, but that doesn't seem to be the case here.

Stalker – Average FPS

The Stalker result is the average of four tests in the freely available "S.T.A.L.K.E.R.: Call of Pripyat" Demo. This engine, the "X-Ray 1.6" is another DirectX 11 capable engine with sophisticated weather effects, detailed texture handling, but only middling render distance. This is the main reason why we're seeing such huge numbers, but it once again gives us an interesting look at the difference in cards. The 5570 is about half of the mid-range cards, but the 5970 stalls again. Based on these results, we can conclude that the dual core CPU is probably holding back the final result a bit, and that the 5970 is again held back. When we go over to the Rampage we'll see things stretch out more expectedly.

Unigine Heaven DirectX 11 Setting FPS

The "Unigine Heaven" test is a freely available DirectX 11 optimized test where tessellation, usually an Nvidia strength, can be adjusted on the fly from none to ridiculous. For this test, I chose a midrange amount, and once again, we're seeing some odd results. The Gigabyte 1156 board seems to have a hard time keeping up with data throughput on the PCI lanes. When we look at the same test on the Rampage, we'll see much more expected results.

Rampage III Extreme motherboard with Core i7 970 running stock ( 4.2Ghz OC results coming in a future test )

Now that we take CPU power out of the equation, we can really let the differences between cards shine. With 1200w of power on tap, and a heavily optimized motherboard, there is nothing stopping these cards from performing at their best, and more importantly, expected difference in overall performance level.

3Dmark 11 Extreme Setting

If you went on 3Dmark alone for your buying decision, you'd think that the 5570 was worthless for gaming, but this is not true at all, and is in fact why I do so many different engines, with both synthetic and real world results.

The two lower and two upper level midrange boards once again compete with each other for fps, but none compare to the sheer muscle of the 5970, even at stock. Although I will publish full results in another article, that 5970 comes in well over 3000 when overclocked to 900/1200. Stay tuned for that.

3Dmark 11 Performance Setting

We expected this. The 5870 and 6870 flatten out on Performance, showing that unless you push the GPUs REALLY hard, you will see almost no real world difference between them. The same holds almost as true for the 6850 and 5850.

Batman Arkham Asylum Benchmark Average FPS

Without PhysX being utilized, there is very little real world difference in Arkham vs. the dual core results. One of the few times we see this. The Unreal Engine 3 is a very clean engine, and runs exceptionally well for how beautiful it truly is.

Just Cause 2 – Dark Tower

The same holds true for Dark Tower. Almost no difference in performance vs. dual core. As we'll see in a moment, that is where that similarity stops.

Just Cause 2 – Desert Sunrise

Remember how the 5970 struggled on these next few tests? Here we see it shine again, pulling a massive lead over the 5870 and showing its dominance and appropriateness with the hexacore CPU. The four middle cards work out roughly the same between runs, and the 5570 struggles to maintain playable framerates, yet it holds on once again, proving that a $70 video card CAN indeed be used with all modern games, so long as you set your games with realistic expectations of Level Of Detail.

Just Cause 2 – Concrete Jungle

The Concrete Jungle, the bane of GPU's on the dual core, proves little challenge to the massively parallel combination of six CPU cores and two GPU cores. The "Middle Four" struggle to maintain 60fps, the modern expectation for smooth gameplay, and the 5570, despite the help of six CPU cores, simply cannot keep up. The fact that we get 21fps at all is really quite remarkable though, considering how much is going on in this test.

Stalker Results – Average FPS

Remember how the 5970 struggled on "S.T.A.L.K.E.R.: Call of Pripyat" and the middle four all evened out? With more CPU breathing room, the results spread out just slightly, but surprisingly, not by much. Of course, at 300+ fps, the difference in cards becomes quite moot, but it's interesting to see that 5570 finally show us some big numbers. It's still a lovely game, and this just goes to show that the 5570 can hang in there with certain engines.

Unigine Heaven DirectX 11 Setting FPS

Perhaps the most interesting test of all; the six core run of Heaven shows us the clearest story of the difference between each of these cards.

The 5970 is way out in front with almost twice the fps of the single chip 5870, which is flatly the same result as the 6870.

The 6850 and 5850 results are identical, while the 5570 struggles to show frames at all. Heaven with DirectX 11 and tessellation is BRUTAL and this final test truly illustrates this point.

Test Result Conclusion:

What we see here is exactly what I was talking about in the beginning. The overall performance of the cards causes them to fall into a ascending rank like this:

Rank of Radeon Card Performance, Ascending:

Radeon 5570 - $70

Radeon 6850 - $200

Radeon 5850- $200

Radeon 6870- $250

Radeon 5870- $300

Radeon 5970- $700

- - -  The following cards will be in my lab next week, and I will review them at that point, but I expect them to fall into place at the expected price point/performance/features spots.

Radeon 6950- $300

Radeon 6970- $370

Sometime in the next couple of months, we expect to see the new top end dual chip Radeon, which is widely expected to be called the 6990, which will be the spiritual successor to the 5970. I hope to have two of them in the lab to test CrossfireX performance for the new top end engines, and possibly do a full Eyefinity Feature, in which I will comprehensively record 1080p video of three monitors running a wide variety of games, with special focus on MMO's. Stay tuned for all of that.

For now, what we take away from the current testing is that the 6x00 cards, once we understand where they lie in the price/performance/features standpoint, are stellar performers, and I can't wait to get the rest of the line on the bench.

Issues?

The Gigabyte 1156 board couldn't handle the 5970 card. I would not recommend this combination, as the 5970 throttles badly.

Other Features

3D Video

Once I get my hands on some active shutter 3D glasses, I can test this feature on an available 3DTV.

Integrity Check

As this is a brand new paradigm in product reviews, I will take a moment to explain the logic behind it in each of my early reviews.

This is how it works:

For my site, my keystone paradigm is “Integrity”. It governs everything that I do. I do not “scoop”, I do not “spy”, nor will I purchase stolen hardware, or less serious but still crappy behavior such as copying content, posting something on my site as if it was my content, with a “via-“ link. There is no via ANYthing on my site. 100% of all of my content is generated by me. Period.

On that same token, I have certain standards for vendors. It has been my long experience that vendors will treat their reviewers equal to how they treat their customers. The only difference between a professional reviewer and an end user is that reviewers don’t pay for the product. However, we still, or at least, I still expect a certain degree of support, communication, acceptance of feedback, and level of behavioral conduct.

As part of this process, I will detail my dealings with each company from first contact, product acquisition, support, and follow up. Exceptional companies will gain a lot of credit here, and the poorly run, greedy, or otherwise “less thans” will be outed.

AMD

First Contact:

As part of the initial research and due diligence for my upcoming MMO Guide and support site, I contacted a number of hardware vendors. Some responded with product, others declined, some promised product but never shipped ( and they too will be outed here as well ) and still others simply wasted my time with a lot of questions and attitude. This is a story about one who not only shipped product, but deserves your money without hesitation.

AMD immediately expressed an interest in my project and after some requisite shuffling and clarification, sent out a set of AMD Radeons, two CPU's, and now the new 6x00 cards. They have continued to show support for this site and my other projects, and I am proud to call them a partner.

Support

AMD has never shown hesitancy to include me in press calls, send me product, or otherwise support my efforts. This is in line with my experience with them over the years as a customer. In stark contrast to competing vendors, AMD is a company I can and will stand behind. End of story.

Conclusion

In this comprehensive review of the Radeon 6x00 series, we've looked at two new cards as compared to their 5x00 series counterparts, gained some new price/performance/features perspective, and delved into the benchmarks for each.

If you are looking for an upgrade today, it's going to depend on what you want to do, and what you currently have. If you have a PSU that only supports one 6 pin SATA connector, get the 6850, which performs almost exactly the same as the 6870 but only requires one connector.

If you want the best, hold off a bit until I release the 6950 and 6970 reviews.

If you absolutely must know if two 6990's are for you, I'll do that too, but since we don't know when we're getting those, any of the current gen cards will serve you quite well. Don't hesitate to support AMD, and game on!

 

 

blog comments powered by Disqus