facebook rss twitter

NVIDIA benchmarks GeForce GTX 480 against AMD's ATI Radeon HD 5870

by Parm Mann on 5 March 2010, 16:19

Tags: GeForce GTX 480, NVIDIA (NASDAQ:NVDA)

Quick Link: HEXUS.net/qawh3

Add to My Vault: x

NVIDIA's next-generation GPU architecture, codenamed Fermi, made a few brief appearances at this week's CeBIT, and though we'll have to wait until March 26 for an official unveiling, the graphics giant has teased its upcoming high-end solution, the GeForce GTX 480, in a surprise YouTube video.

The clip, featuring NVIDIA's director of technical marketing Tom Petersen, shows a GeForce GTX 480 running in a high-end Intel X58 system equipped with a Core i7 960 processor. After a brief explanation of DirectX 11, particularly the goodness on offer from hardware tessellation, Petersen, who ran through 'Fermi's' features in a HEXUS.tv show, provides a performance comparison between the GeForce GTX 480 and AMD's ATI Radeon HD 5870 using the Unigine 'Heaven' benchmark.

Of course, any benchmark originating from a manufacturer needs to be taken with a pinch of salt, but NVIDIA's graph suggests the GeForce GTX 480 is superior - that is, at least when heavy tessellation comes into play.

What's interesting is that most users may have thought the GeForce GTX 480 would dethrone the current king - AMD's ATI Radeon HD 5970 - as the fastest single graphics card available. If this evidence is anything to go by, those beliefs may be cast into doubt - as the GeForce GTX 480 looks to be ahead of a Radeon HD 5870 when tessellation is heavily utilised, but almost on par at other times.

We won't doubt a card with some three billion transistors, but to say we're eager to get it into our labs to see what it really can do would be an understatement.

The entire video clip, which also includes a brief look at GeForce 3D Vision Surround, can be viewed below.



HEXUS Forums :: 51 Comments

Login with Forum Account

Don't have an account? Register today!
OMG, this is now really worrying for nVidia….this screams of them hand-picking the one benchmark that shows it in a good light vs the 5870 and even then, it's marginal?

You will be able to count the amount of nVidia fanboys in the world by Fermi sales…….
Assuming those figures are anywhere near accurate, it suggests the 480 can maintain double the minimum fps during intensive scenes. This is much more important than getting 200fps while looking at the floor.
sadbuttrue
Assuming those figures are anywhere near accurate, it suggests the 480 can maintain double the minimum fps during intensive scenes. This is much more important than getting 200fps while looking at the floor.
You mean ‘during intensive benchmarks that don’t represent real world games'?

Benchmarks absolutely have their role, and it suggests that the tessellation in nVidia's card really is quite good.. but given it's a dx11 only feature I have my severe doubts that developers are going to go anywhere near a tessellation limited game - at most it'll be a bit of eye candy ontop of an engine really designed for consoles. Raw polygon and texture throughput, together with shader speed, are going to be the most important factors in whether this is better at gaming or not, and cherry picking one specific facet suggests they're struggling to even beat ATI's upper rangecard that's been out for a while, let alone ATI's top card.
sadbuttrue
Assuming those figures are anywhere near accurate, it suggests the 480 can maintain double the minimum fps during intensive scenes. This is much more important than getting 200fps while looking at the floor.

except those intense scenes are tessellation heavy where in 10.2 and below the 5870 sucks, however the 10.3 beta seems to have fixed this and now the 480 should be only marginally better at this benchmark, so it may still be crap in other benches/games.
So nVidia find that their own benchmark gives their cards better scores? Yeah, I'll wait for the 3rd party results to come out, thanks.

Even if that is the result, I bet the nVidia cards will require more juice, run hotter, and have a higher failure rate.