Reviews - Inno3D GeForce GTX 260 OC vs. Gigabyte Radeon HD 4870: the aftermath

[HEXUS]

Long overdue upgrade? We put a pair of sub-£250 graphics cards head-to-head.

Read more.

[kalniel]

At 1,920x1,200, the GIGABYTE Radeon HD 4870 1GB performs some 25 per cent slower than Inno3D's GeForce GTX 280 OC. Note, also, that this is using AMD's yet-to-be-publicly-released Catalyst 9.1 beta drivers.

Where's hurleybird to eat his words?

[Kerotan]

Damn, wish I had a spare £250 lying around, wonder what the market will be like around August, should finally have some money to burn then

[Zak33]

my initial musings on this review are varied.

A: Bloody good comparison guy's bang on the money as ever.

B: CoD4 players will be GAGGING for a 4850x2, so these two are tricky pricing

C: If I were Zalman I'd be a tad embarrased about a company making my product so... loud! Zalman were the Guru's of Quiet and Efficient. I know that heatsink does a good job, but this whole ATI 4850/4870 fan speed thing has drawn on for too long now. Without a review like this, how woud you know whether you're gonna be buying an ear-splitter or not? I've got some experience with 260's and they are a tad warm, but they're quiet and dont seem to mind being warm.

D: Power drain at tick-over, ie not gaming, and the nVidia boys have blown that ATI set up out of the water. Many many gaming rigs are used for surfing/homework/video too. To have a high power drawn is just plain wrong. 4870 uses 34w more at idle that a 260

40 hours work (non gaming) per week, 52 weeks per year = 2080 hours x 34watts more than a 260 = 70kwh per year. Paying around 15 p per kwh = £10 per year extra in electricity at tickover , PLUS the xtra gaming draw over a 260. Might as well save that tenner and spend it on something else!

[twobombs]

All right, lemme see here.... Overclocked products are always sold with a premium. Real enthousiasts' buy the vanilla thing and OCL it, as most OCL products are reflashed vanilla versions, apart from the reflash no real difference with the vanilla product. So the OCL moniker is extra change for the retailer.

Second: I find it strange that in a lot of reviews the horsepower of a card is measured with a game instead of a real GPGPU application that very much can compare apples ( AMD ) with apples ( NV ) on a GPU level. We are benching a GPU, right ? Not a game.

So with this in mind I would like to recommend all reviewers to use programs like barsfw to evenly measure CPUs and GPU combos, wattage and reflect the outcome of those programs to games so you can see where the cpu/gpu/ppu ratio kicks in or out.

Some PhysX comparison would be fine with me also... hi AMD

Thank you for your time.

[Tidus]

The 4850 X2 is a good looking card at this price point and looking likely to be my next card, I stumbled accross a Sapphire ATI Radeon HD 4870 1024MB for £166.74 inc VAT at overclockers if anyone is interested, it's on a this week only deal so today may be the last day but a big saving over the Gigabyte version.

Second: I find it strange that in a lot of reviews the horsepower of a card is measured with a game instead of a real GPGPU application that very much can compare apples ( AMD ) with apples ( NV ) on a GPU level. We are benching a GPU, right ? Not a game.

So with this in mind I would like to recommend all reviewers to use programs like barsfw to evenly measure CPUs and GPU combos, wattage and reflect the outcome of those programs to games so you can see where the cpu/gpu/ppu ratio kicks in or out.

Some PhysX comparison would be fine with me also... hi AMD

Thank you for your time.

People buy expensive graphics cards for games, so I can't see anything better than a game to guage it's performance, If anything I think Hexus should add a few more games in to vary the results more.

By barsfw are you talking about a program called BarsWF?, to my knowledge (after a quick Google search) this program is a MD5 cracking tool that utilises the GPU power?, If so what use is it to the average user in a card review and does it run on AMD hardware or just Nvidia?

[twobombs]

It runs on both NV & ATI and can reliably be used to gauge the raw GPU shader performance of a card, something that often is masked by pure fps ( frames per second ) gaming benchmarks. Other programs also look at raw GPU performance such as the GPU client of the folding@home project.

With the benchmarks of those programs in mind one can easily see which games are cpu bound, and which are GPU bound whatever SKU is being thrown at the benchmarking table.

Besides that, games scale tremendously in things like PhysX support ( mirrors' edge, UT3 ) or multicores. So even a pennywise dog+pony game benchmark that I see before me here today, a lot of benchmarks can be pretty much be (in)validated by a runup that is featurewise coupled with some raw GPU numbers; CPU usage during benches can also be very interesting, enlightening.

Or do I sound too much like an NV fanboi, as that could very well be the case.