Technology Primer - ATI Radeon HD 6000-series

Written on October 22, 2010 by William George
Share:

Introduction

ATI has just launched their next-generation graphics card line, starting the process that will see the venerable Radeon HD 5000 cards replaced by 6000 models over the coming months. The first such cards, released today, are the 6850 and 6870 - known also by their codename ‘Barts’.

Radeon HD 6850 & 6870 Details

Performance-wise these cards are very similar to the previous generation units with similar model number (5850 and 5870). The big difference is that these new cards put off less heat and come in at a lower price point, plus they sport some very nice new features. For example, the HDMI support on these cards is revision 1.4a - that brings higher resolutions and refresh rates, similar to dual-link DVI, and can enable other technologies like 3D Blu-ray movies to be used. These cards also support ATIÂ’s Eyefinity technology, as did their predecessors, which allows multiple monitors to run off a single card for either greater immersion in games or increased productivity.

The lower heat output from these cards, compared to the previous generation, means the cards can in theory run quieter. WeÂ’ve observed that with the samples weÂ’ve been testing here over the past week, so we can vouch for the reference cooler design from ATI as being a good option in that regard. It is likely that various graphics card manufacturers will install customized cooling solutions as well, but here at Puget Systems weÂ’ll be on the lookout for the quietest and most reliable options rather than heatsinks and fans that simply look cool.

Lastly, the fact that ATI is aiming these new cards at lower price points means they will compete with different nVidia cards than the models they are replacing. That has already put some pressure on nVidia to lower prices, so it will be interesting to see what changes they make to their already strong lineup of GeForce 400-series cards in order to adapt and match the price-to-performance ratio of these new models.

Performance Tests

For those interested in performance numbers, we sent a couple of our production techs home with the sample cards to do real-world testing on these cards. Both techs we selected were already running ATI 5000-series cards, so we can see how these new models stack up against them, plus we gave them GeForce GTX 460 768MB samples to test as well.

Tester #1, John, had been using a Radeon 5770. He took home a 6850 to play with, which should be at about the same price-point as the 5770 but in theory offers better performance. Here are his results, measured in frames-per-second (FPS):

Benchmark TestRadeon HD 5770 1GBRadeon HD 6850 1GBGeForce GTX 460 768MB
Dirt 2, DX11 @ 1920x1080, High settings,
2x MSAA, Average FPS
47.260.364.4
Dirt 2, DX11 @ 1920x1080, High settings,
2x MSAA, Minimum FPS
38.147.650.5
Dirt 2, DX11 @ 1920x1080, Ultra settings,
8x MSAA, Average FPS
44.357.153.9
Dirt 2, DX11 @ 1920x1080, Ultra settings,
8x MSAA, Minimum FPS
35.045.042.6
Just Cause 2 @ 1920x1080, Max detail, 8x AA, Dark Tower-37.8231.04
Just Cause 2 @ 1920x1080, Max detail, 8x AA, Desert Sunrise-47.0744.35
Just Cause 2 @ 1920x1080, Max detail, 8x AA, Concrete Jungle-28.4826.04


From JohnÂ’s benchmark results it can be seen that the 6850 definitely performs better than the 5770, so if the market doesnÂ’t cause the price to soar over ATIÂ’s recommendation then it will be a great value. The difference was big enough - about a 30% increase - that he didnÂ’t bother including the 5770 in the second round of testing, but the GTX 460 comparisons here prove a little more interesting. The GTX 460 is nVidiaÂ’s card with the closest MSRP to the new ATI card, and performance between the two is very close. It looks like nVidia leads slightly with lower anti-aliasing settings, but when those are cranked up the 6850 takes the lead. Even then, though, both cards perform very close to each other - so either would make an excellent purchase.

Tester #2Â’s name is Ruben, and he started off with a more powerful computer and video card: the ATI Radeon HD 5870 Eyefinity Edition with 2GB of RAM. For testing, we gave him the new HD 6870 1GB - which based purely on specifications should perform around the same level, but cost less; lets see if that holds true:

Benchmark TestRadeon HD 5870 2GBRadeon HD 6870 1GBGeForce GTX 460 768MB
Dirt 2 Demo, DX9 @ 1920x1080, Ultra settings,
8x MSAA, Average FPS
77.270.761.3
Dirt 2 Demo, DX9 @ 1920x1080, Ultra settings,
8x MSAA, Minimum FPS
63.454.853.4
Just Cause 2 @ 1920x1080, Max details, 8x MSAA, Concrete Jungle36.2433.4726.08
Unigine Heaven @ 1920x1080, Max settings, Peak FPS40.234.729.5
Unigine Heaven @ 1920x1080, Max settings, Average FPS15.814.710.9
Unigine Heaven @ 1920x1080, Max settings, Minimum FPS6.67.62.2


Based on these numbers it looks like the 6870 does indeed come close to the performance of the 5870, but the older card still has an edge in most tests - maybe in part due to this particular sample card having more onboard RAM. Since the 6870 is expected to come out at a much lower price, though, the new card should be a very good value. You can also see that it beats the GTX 460, though in terms of price/performance ratio the more affordable nVidia card holds its own.

Conclusion

Adding features and cutting prices are both great moves in terms of advancing computer technology, so here at Puget Systems weÂ’ve been anticipating this launch for some time. Increased competition is also great, as it spurs all involved to even better products!

wargasm

Im a little confused as to how your tests varied so much from Maximum PC's results. While MaxPC is a reputable source of benchmarks and reviews, I think your results may be suspect. Could you please clarify your process and methods used?

Posted on 2005-12-01 07:08:44

There is no variation in test results that I know of. We ran the same tests and got the same results. The only difference is in the comparisons and the iterpretation of the results. In fact, I don't think Maximum PC even provided raw scores...they just gave "comparison scores" to their zero point configuration, which was a FX55 6800 Ultra SLI machine.

So the difference: Maximum PC compared our system to a FX55 6800 Ultra SLI machine and said we were slow at office tasks, but that we kept up in gaming. In our tests, we compared to 6 other machines that run 7800GTX video cards (ie, setting the video card as a constant), and looked at how changing the processor affected the performance, and found that Pentium M came out on top, followed by FX57 and 4800+ dual core.

Posted on 2005-12-01 09:24:36
wargasm

"With its single 7800 GTX videocard, the Puget system couldn’t compete with the big boys in our game benchmarks." This is the line that mostly caught my attention after reading your benchmarks and information. Your results seem to show the PentM clearly beating out the game heavy performing FX-57.

In the least I have to say it is a very interesting read, I had just never thought of the PentM core as a gaming use processor (perhaps im just a bit behind on the times). I wish I still had the ability to run these processors side by side in a like enviorment.

One question, to what degree of advantage do you think the Raptor HD and the DDR2 memory played?

Posted on 2005-12-01 10:48:22

Right, they say "with a single 7800GTX"...ie, in that statement they are comparing to SLI 7800's. We're holding the video card constant in our comparisons so that we're looking at the difference the CPU makes.

I think we'd need to run more comparisons to answer your quetion about RAM and hard drive. We used benchmarks that included a Raptor as much as possible in our comparisons, but a few systems didn't have that. My feeling is the hard drive wouldn't play much role. It would make the levels load faster, but not gameplay. Memory would probably play a bigger role, but I haven't run any comparisons on that. Can't easily build a FX57 with PC2-5400 memory :)

Posted on 2005-12-01 11:00:39
wargasm

Always great to see knowledgeable staff willing to discuss thier products with thier consumers. Best of luck to your staff and your company, a great example of how business should be done :).

Posted on 2005-12-01 11:24:08
jimlickster

How do you explain the $4700 Price tag though. This Machine would cost MAX $2500 Anywhere else.

Posted on 2005-12-02 09:48:16

I'd love to see a link to a system like this for $2500 :)

Here is a breakdown of the pricing:
http://www.pugetsystems.com/sa...

Keep in mind that we built it when the CPU and vdieo card had just come out, so pricing has certainly dropped since then...just not to $2500. Our price today on a system like this would be more in the $4000 range.

I doubt anyone would be interested in this exact system though -- this system was a demonstration of a passive radiator and a demonstration of how well a Pentium M does in gaming. It will be more common for us to see this liquid cooling system applied to a higher performance setup, and to see the Pentium M applied in lower end gaming systems (HTPCs, SFF PCs, etc).

Posted on 2005-12-02 09:55:23
Chumly

I had a long drawn out post, but I'd like to just add my opinion on PC's. Choose the right processor for applications, then choose the right video card(s) for gaming. A cheaper 650 would have done better in apps, and that saved $400 would allow a pair of 7800GT's. Then you'd have better benches in both apps and games over the reviewed system. Afterall, You can watercool any proc and run silently.

For $4700, the customer should just get more than a "demonstration" or experiment, hence the review ding. Even at the lower $4K estimate, I'd ask for more than an M, a single card, and 62Gigs of space.

Posted on 2005-12-02 23:55:17

I think the confusion developed when it was suggested that we try something different and unique, so we did. But you're absolutely correct in saying that MaximumPC magazine is just about customers looking for systems to buy.

By the very nature of that premise, MaximumPC magazine wants to compare similar systems, targetted at their aduience. There's not anything wrong with that. I just think there was a little disconnect between what we thought they'd think was a neat system, and what they were really looking for.

To be honest, I don't think the "submit a system for review" method is really good at showcasing the kind of customization we do. By definition, a submitted system is static, and looked at individually. We offer dynamic customizability, which gives the control to the consumers. For example, MaximumPC magazine wrote that they were dissapointed by the small size of a single Raptor Hard drive. While it's true that some people would prefer more space, others wouldn't. It's a valid complaint that our submitted system had small storage space, but it's not really a "real-life" problem any customer would encounter, since they'd just select a larger hard drive to begin with.

So, in our defense, I think a lot of the reasons our score was knocked down were fairly legit conceptual problems, but nothing that any customer would actually ever encounter.

Next time we know!
:thumbsup

Posted on 2005-12-05 14:32:47
Filidor

I really enjoy these types of articles guys. Thank you and keep up the good work!

Posted on 2010-11-06 08:39:40

I'm glad you like them! We've got another coming soon on power supplies - explaining wattage and efficiency. Should be up sometime next week hopefully :)

Posted on 2010-11-06 12:35:05