The Radeon HD 4870 1GB sure dropped to $150 pretty quickly, didn’t it? The Radeon HD 4890 really isn’t all that far behind at $190 (as low as $170 with mail-in rebates). So, for the Radeon HD 5850 to be a success at $259, it’d better be appreciably quicker, right?
Nvidia has its own high-end bruisers around the same price range, too. A GeForce GTX 275 at $210 is mighty tasty. And a GTX 285—the company’s fastest single-GPU board available—isn’t bad at $330 or so given its single-GPU flagship status (less than $300, after some of those rebates).
If you haven’t yet checked out our review of the Radeon HD 5870, you might want to give it a quick peek. After all, the Radeon HD 5850 under our microscope today centers on the same fundamental architecture as that board (and I don’t think I can swing another 10,000 word story this week, so this piece isn't going to cover all of the GPU nuances).
Answering The Lynnfield Question
But that doesn’t mean we can’t break some new ground. One of the criticisms I saw come up in the comments section was that we used a $1,000 processor overclocked to 4 GHz for testing AMD’s Radeon HD 5870. Of course, that configuration was by design. These new GPUs are so powerful that we wanted to give them as much room to “breathe” as possible, without seeing congestion in the benchmarks due to processor bottlenecks. This presents a bit of a theoretical question to the folks running Core i7, LGA 1156-based Core i7s, a Core 2-series chip on P45, or 790GX. Mainly, does the move from a x16 PCI Express 2.0 slot to a x8 connection affect the performance of such a powerful GPU.
In order to help answer that, we took our Core i7-870 and overclocked it to 4 GHz on Asus P7P55D Premium. I suspect that in a Lynnfield-based configuration, the Radeon HD 5870 will be a less-popular choice than the cheaper Radeon HD 5850, so we tested a pair of 5850s on both Intel-based platforms to shed a little light on this one.
We also dropped the Core i7-870 to its stock speeds in order to isolate the effect of processor performance versus our overclocked Lynnfield-based results.
Is CrossFire Worth It?
When ATI launched the Radeon HD 4770 at $110, we couldn’t pass up the opportunity to compare two of the cards together against Radeon HD 4890s and GeForce GTX 275s. But the Radeon HD 5850 is not a cheap piece of hardware at $259. A pair costs just under $520. The only single card in that neighborhood is a GeForce GTX 295, which can be found for roughly $500 and hasn’t yet been discounted, despite AMD’s Cypress launch. To be fair, there isn’t yet a need, as the 5870s are still in extremely limited supply, so the challenge seems to be getting your hands on one.
But maybe the Radeon HD 5850 will change that. Today, we’ll be looking for a single Radeon HD 5850 to stand up to Nvidia’s GeForce GTX 285—a card AMD couldn’t contend with using a single-GPU solution in the past—and a pair of 5850s to at least eke past the GeForce GTX 295, priced similarly.
- Introduction
- The Makings Of Radeon HD 5850
- Hardware And Benchmark Setup
- Benchmark Results: 3DMark Vantage
- Benchmark Results: S.T.A.L.K.E.R.: Clear Sky
- Benchmark Results: Crysis
- Benchmark Results: Far Cry 2
- Benchmark Results: Left 4 Dead
- Benchmark Results: World In Conflict
- Benchmark Results: H.A.W.X.
- Benchmark Results: Resident Evil 5
- Benchmark Results: Grand Theft Auto IV
- Power Consumption And Noise
- The Lynnfield Element
- Conclusion
As long as ATI continues to make high-performance graphics cards which can still fit sensibly (9.5") inside smaller pc cases, Nvidia are sure to lose the technology war. I certainly hope Nvidia aren't planning any 19" long cards!
Crysis @ 1680 @ High @ 50 frames w/ 1 gfx card = how it should be.
Yet it won't be as long as game developers create engines that are capable of outstripping the capabilities of single GPUs pushing the hardcore gamer into multi-GPU setups. But given that GPU manufacturers now have viable multi-GPU technology they will push that hard to sell more cards, thus allowing game developers to keep pushing the upper limits of their engines.
Now if the hardcore gamer stopped buying muti-GPU systems then the above cycle will end, but they won't - either through pride or stupidity the hardcore gamer will say "but to run this at maximum I need 3 GTX-whatevers" and will go and buy them.
In a couple of years single cards WILL be able to play Crysis as you say - there were many similar complaints about Elder Scrolls' requirements when it came out, but it's hardly an issue now - abut frankly I'm tired of people using Crysis as that measuring stick - it's totally unrealistic! Either through optimisim or pure stupidity, the Crytek engine used is just beyond any sensible levels of technology we have at the moment. I thought Doom 3 on Ultra quality was a bad idea when it first came out and then tech caught up, but I'm not sure we'll see that with Crysis any time soon.
I don't think it was tested at maximum resolution and effects. They probably had cutting-edge hardware, built so it was playable at maximum settings and then turned up the dials of what the engine could do visually - at that point you only need to do stills to test the effects, or pre-render sequences as video to test it out.
It's not hype, it's basic maths: if you have something that goes X fast but is capable of going Y faster then pushing it to Y will get things done faster than X.
Doesn't matter if it's games, video encoding, finance calculations or anything - if your chip can be overclocked to get the job done faster then why not do it?
Also don't think that games can't CPU intensive: Crysis for instance has a purely software-driven sound engine that will be processed by the CPU, and GTA IV is such a bad port it runs even a quad-core really hard.
The real discussion though is dual-core vs quad-core in games - do you go quad core when most games won't take advantage of it? For the longest time the consensus was always a faster dual-core was better for you, but GTA IV and a few others have now started being quad-aware and that trend is no doubt growing.
Usually it's massive voltage increases and/or not dealing with heat properly that causes damage to the CPU - the D0 stepping i7 for instance only needs a tiny voltage boost, if at all, to reach the 3.8GHz+ mark, and keeping the temps to around 75 degrees or less keeps the chip OK. So that being said, overclocking the CPU to these levels won't do it any harm
The reason for the overclock though in this case is to remove any possible CPU bottlenecks when driving such a powerful Crossfire setup - as the article is about the GPU's performance you want the numbers coming out to be as pure as possible.
Real-world? Overclocking isn't going to hurt you and if you can and keep it under control, why the hell not? But if you can get the perceived standard 60fps+ out of a CPU at stock then yeah you're probably just splitting hairs or showing off by cranking your CPU up for those extra 20 or 30fps you're never going to see.
Applications is a slightly different kettle of fish IMO. I do a lot of graphic and video work and I render and encode masses of data. As a result I can never go too fast, and even if I get an extra 10% performance by overclocking I'll take it - 10% off a 60 second render may only be 6 seconds, but if I can shave 24 minutes off a 4-hour render then it's worth it.
great choice of CPU and that one needs to be OC
in the past after OC my system i have seen an increase in 30% for 3D mark.
as long as ur temps and voltage's are fine theres no down side to OC. just dont be tempted to push the CPU to its max in order to obtain that last 2-5% performance boost lol
Or is the sheer mention of the green team in a red team article really that much of an insult?