New GTX260 vs 4870 1GB

[BenSkywalker]

So you're saying the 4870 WITH AF is faster than the 280 without (at least in CoD)?

Huh? 4870 with AF is 69.9, 280 without is 90.9.

Also, you know for sure that the testing methodology is the same between the 2 sites?

They list there testing methodology before each review. It differs between given reviews for either site, but listing graphics settings used it standard for both of them.

I know it doesn't say it in the chart but AFAIK that's a setting they use for their vid card reviews isn't it?

When they use AF they make note of it.

 

[thilanliyan]

Originally posted by: Wreckage
Are they using the same settings, same CPU, same motherboard, same drivers......

There's more to benchmarking a video card than just the GPU and the game.

Which is exactly what I was saying in the first place. I said it depends on which site you read. I know different sites use different methods...which is why I'm saying you can't just take ONE site as proof.

Originally posted by: BenSkywalker

So you're saying the 4870 WITH AF is faster than the 280 without (at least in CoD)?

Huh? 4870 with AF is 69.9, 280 without is 90.9.

http://techreport.com/r.x/radeon-hd-4870/cod4-1920.gif
It shows the 4870 beating the 280 there. The question I was asking is whether you think the 4870 with AF is faster than the 280 in the same test, compared to another test without AF on both cards? (ie. Isn't the 280/260 supposed to take a very minimal performance hit with AF compared to the 4XXX cards) (IIRC it was like that when they introduced G80)

 

[BenSkywalker]

What I meant was the 4870 with AF is faster than the 280 in the same test, compared to another test without AF on both cards?

Yes, ATi is doing a LOT less work with AF enabled then nV is for the same setting. Same thing with 8x AA between the two companies, although it is a lot less obvious with AA then it is with AF where the difference in IQ is rather pronounced.

 

[apoppin]

The question I was asking is whether you think the 4870 with AF is faster than the 280 in the same test, compared to another test without AF on both cards? (ie. Isn't the 280/260 supposed to take a very minimal performance hit with AF compared to the 4XXX cards) (IIRC it was like that when they introduced G80)

it depends on the game being tested. Rarely is the 4870 faster than the 280 at the same settings.

However, the 48X0 series appears to have less of a performance penalty enabling 16xAF - percentagewise - than Tesla at similar settings at 19x12 and 16x10 in my experience.

 

[thilanliyan]

Originally posted by: BenSkywalker

What I meant was the 4870 with AF is faster than the 280 in the same test, compared to another test without AF on both cards?

Yes, ATi is doing a LOT less work with AF enabled then nV is for the same setting. Same thing with 8x AA between the two companies, although it is a lot less obvious with AA then it is with AF where the difference in IQ is rather pronounced.

I didn't know that there was such a large difference (I think I have read that nV cards starting from G80 do nearly perfect AF but I didn't know that the AMD cards are doing a "LOT" less work). Got any links I can read?

Also, can anyone else confirm that GT200 cards take a large performance hit from going to 16xAF?

 

[BenSkywalker]

I didn't know that there was such a large difference (I think I have read that nV cards starting from G80 do nearly perfect AF but I didn't know that the AMD cards are doing a "LOT" less work). Got any links I can read?

Links no, I just have the hardware to look at I know BFG was talking about in his thread when he got his 4850 too, it really is quite obvious just by looking at it(ATi is using a very rough approximation for their AF, nothing like nV does).

Also, can anyone else confirm that GT200 cards take a large performance hit from going to 16xAF?

Not large, larger The nV2x took a large performance hit(60%-70% was normal).

 

[BFG10K]

Originally posted by: BenSkywalker

Yes, ATi is doing a LOT less work with AF enabled then nV is for the same setting. Same thing with 8x AA between the two companies, although it is a lot less obvious with AA then it is with AF where the difference in IQ is rather pronounced.

With AF I agree completely, but not with 8xAA. Both use a sparse grid and take a genuine 8 multi-samples so overall both the workload and IQ should be identical.

 

[dug777]

If I wasn't so lazy I'd do an AF comparison between my 4850 and my better half's 9600GT.

Alas, I'm too lazy

 

[apoppin]

Originally posted by: dug777
If I wasn't so lazy I'd do an AF comparison between my 4850 and my better half's 9600GT.

Alas, I'm too lazy

BFG and i already did a Crysis IQ comparison with 8800ultra and 4870; i guess he is finishing it up .. after all ..
- and i am lazy as heck ..

they are pretty close but i would also give the edge to Nvidia - especially with AF [which takes less of a performance hit on AMD cards, at least in Crysis]

i have also just started on a full IQ comparison between Tesla and r770 .. tentatively we are using Crysis, CoJ and Oblivion and doing the full range of tests with both Dx9 and Dx10
- next month, i believe it will be done; these take quite awhile to do