AnandTech GPU Bench

[Mr Krugga]

The 5800 series definitely has an efficiency issue as performance increase over the 4800 series simply does not match the higher specs. A good example is the performance difference between the 4890 and 5870. I'd expect something near double the performance, but in reality it's nowhere near that.

I've got a feeling that the R600 architecture is slowly maxing out.

You're so wrong now. First of all, these results are here for comparison only, you cannot calculate performance difference judging by a few games. Second of all, you should actually compare Radeon 4870 to 5870 (ATI never decided to make 5890, rather jumping straight onto 6000 series ship). And GeForce 470 to 275.

Here are the results:

http://www.anandtech.com/bench/Product/176?vs=162
http://www.anandtech.com/bench/Product/169?vs=160

 

[RavenXXX2]

5870 kills a 4870 in the latest DX10 games like BC2, metro 2033, easily over 100% performance increase with max IQ.

 

[Freddie1980]

So just how much better is the HD5850 over the GTX460 1Gb??

There's your answer, based over 10 games, 3 different resolutions so any abnormal results are ironed out (not that there are any), the Radeon has an overall 11% lead over the GTX460 1Gb and uses less power to do it.


And just for a bit of fun I thought we could compare the HD5850, GTX460 1Gb, HD5830 and HD5770 together and see what the differences are.

*** Note there were no 2560x1600 benchmarks for the HD5830 or HD5770 so I took those results out for the HD5850 and GTX460 to give all cards an even platform.

The GTX460 has quite a big lead over the HD5830 and maybe I'll include the 768mb model to see what the differences are between it and it's full beef variant.

 

[Marine-RX179]

Freddie1980 can you do a compare for between the 5850 and GTX470? The results from AnandTech seem to indicate the gap between 5850 and GTX470 is wider than the gap between the GTX460 1GB and 5850...would be like to see how much faster is the GTX470 in % with that method of calculation.

 

[Mr Krugga]

Lol Freddie, been doing the same thing for some time now, here are the results :

Now with Radeon 5850 at £205 ($270 in the US) and GeForce 460 1GB going for £165 ($230), there is 24% extra (17.5% in the US) you pay for 10.5% performance increase (Freddie claims 11%, Techpowerup 12%, etc.). The results in other threads are also pretty consistent thus it becomes harder to recommend Radeon 5850 over GeForce 460 1GB these days (until ATI cards back their prices a little).

 

[titaniumx3]

How much faster is the GTX460 than a GTX285? As I recall, that was fairly close to a 5850.

 

[Mr Krugga]

How much faster is the GTX460 than a GTX285? As I recall, that was fairly close to a 5850.

Not faster at all, just about the same performance really. Lacking DirectX 11 that is.

 

[555BUK]

According to Gibbo, ATi's cards are still very much in really high demand.

A salesman is hardly going to say don't buy our stock - it's overpriced, is he? Keep both ATI and NVidia fanboys content and more stock gets sold. More sales equals bigger commision, equals newer and faster car...

 

[kylew]

A salesman is hardly going to say don't buy our stock - it's overpriced, is he? Keep both ATI and NVidia fanboys content and more stock gets sold. More sales equals bigger commision, equals newer and faster car...

I understand that, but the prices of them have held steadily since release. High demand can be the only real reason, considering the prices actually went up post release says a lot as well.

They're still overpriced of course, but are clearly selling very well. GTX470s and 480s are a different matter though, the 470s especially considering how dramatically their prices have dropped in a matter of weeks.

 

[Arcane]

No 470 SLI results on the bench, strange.

 

[Freddie1980]

Lol Freddie, been doing the same thing for some time now, here are the results :

The 0.5% difference is because I tend to round up concluding percentages I think the actually number was about 10.67% - when I was taught percentages and rounding the rule is if it’s .5 or above you round up and if it’s below you round down. The fact that we have got practically the same results is helpful as it shows I haven’t made bobo which could be embarrassing.

BTW it's not Freddie is saying 11% it's the guys at Anandtech, I'm just the numbers into a helpful list with actual % differences between them. I'm not actually thinking it will more helpful to report the performance differences at different the resolutions. At 1680x1050 the GTX460 holds up well against the HD5850 but above that the HD5850 flexes it muscles this way people can make a buying decision based on what monitor they have.

Freddie1980 can you do a compare for between the 5850 and GTX470? The results from AnandTech seem to indicate the gap between 5850 and GTX470 is wider than the gap between the GTX460 1GB and 5850...would be like to see how much faster is the GTX470 in % with that method of calculation.

I've already done that comparison (see here). Just bear in mind that the data in that chart was when Fermi came out in April/May or whenever it was so drivers may have made some changes so I'll do it again this evening to see if there is any significant difference but given the games involved in the benchmark the impact of drivers on single cards will be minimal.

---------------------

I've just has another look at the numbers to compare how well cards turn electric into frame rates. I think as we all suspect the HD5850 is an efficient ******* and requires just 4.44 watts for every FPS (based on the aggregate results posted above), 2nd, 3rd and 4th places are more close but the maybe somewhat surprisingly the GTX460 comes 2nd on efficiency requiring 5.2 watts per FPS beating out the HD5770 that needs 5.47 watts and the black sheep of the Radeon family the HD5830 needs 5.95 watts.

 

[straxusii]

Great find, it's really missing a summary though.

I would want to see at the very top a summary of the contest:

Number of benchmark wins, eg 3 vs 5
% performance, eg 5870 100% GTX480 110%

 

[Mr Krugga]

Great find, it's really missing a summary though.

I would want to see at the very top a summary of the contest:

Number of benchmark wins, eg 3 vs 5
% performance, eg 5870 100% GTX480 110%

There can't be a summary with so little data tbh, it would be unfair even though we know AnandTech isn't biased at all (or so to speak), the benchmarks are truly limited and are here for comparison only. We would need to collect the data from many sources, perhaps even individuals, to assume the real differences between specific models. Now we have something to base our opinions on, we can fulfil the data if needed and such a site is needed for sure, but cannot be taken as a definite.

 

[WingZero30]

Wow!! 5970 almost twice as fast as a 5850. Also it seems that in general 470 gives on average 10 fps more than 5850 but drawing almost 100W!! more.

 

[Marine-RX179]

Also it seems that in general 470 gives on average 10 fps more than 5850 but drawing almost 100W!! more.

The extra 100W during gaming probably won't even cost more than £10~£20 extra a year on the on the energy bill though. Plenty of people using poor efficiency PSU that could easily draw that much more power from from the wall sockets

 

[straxusii]

There can't be a summary with so little data tbh, it would be unfair even though we know AnandTech isn't biased at all (or so to speak), the benchmarks are truly limited and are here for comparison only. We would need to collect the data from many sources, perhaps even individuals, to assume the real differences between specific models. Now we have something to base our opinions on, we can fulfil the data if needed and such a site is needed for sure, but cannot be taken as a definite.

Yes but I'm a simpleton and like to see X scores 9, Y scores 10

 

[MoRT489]

That's a great tool especially the comparison part. It could do with a few more games (e.g. Metro 2033, Far Cry 2, COD). Would be great if you could choose the resolution as well.

 

[Mr Krugga]

Bumping the thread with some more info.

Drivers they use and the rig for testing:

1. Nvidia ForceWare 258.80 - GeForce 460 768MB/1GB/SLI/465/470/480/SLI
2. Nvidia ForceWare 197.13 - GeForce 8800GT/260/275/285/SLI/295
3. AMD Catalyst 10.7 - Radeon 5850/5870/CF/5970
4. AMD Catalyst 10.3a - Radeon 3870/4850/4870/4890/4870X2/5750/5770/5830

• Intel Core i7 [email protected]
• Asus Rampage Extreme II/BIOS Rev. 1504
• 3x2GB Patriot Viper 1333MHz/7-7-7-20
  
  
• Antec Quatro 1200
• Thermaltake Spedo
  
  
• Windows 7 Ultimate 64-bit
• Intel 9.1.1.1015 Chipset Drivers

 

[Mr Krugga]

Another set of results, GeForce 470 vs Radeon 5850: