5850 vs GTX470?

[tosgamer]

1) I happen to prefer real results to canned benchmarks, and if you check [H]'s reviews the 5850 actually WINS in performance

I'll take canned benchmarks any day over "real results". Canned benchmarks are chosen by the developers to represent how the game will play overall. They are always the same with no variables. I much prefer that over someone just running around for 10 minutes. There are way too many variables to fairly compare 2 cards in "real results" for me. For example in metro 2033 there are some areas where just looking two inches higher towards the open sky will drop down 6-7 frames compared to if a player looks dead ahead. Am I suppose to believe the player was able to retrace his steps 100% right down to the camera position through each play through?

2) Even other canned sites disagree with that: http://techpowerup.com/reviews/Palit/GeForce_GTX_470_Dual/29.html

There is less than a 10% performance difference at all resolutions, and at 2560x1600 it drops to 1%. 20%? Not a chance.

Nice one using an article from May. Its not 20% but it is a very consistent 10% right now. Here is the latest chart from them showing a 9%-13% difference between the 2 cards at every resolution. Strangely enough there is now a 13% gap for your 2560x1600 resolution. http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTS_450_SLI/23.html Magic drivers? Maybe not but a definite improvement.

3) You want to talk about the effect of new drivers and then link to a review using old drivers for the ATI cards? Nice, real nice. But for what it's worth, [H] did a comparison of Fermi's new drivers vs. the launch ones: http://hardocp.com/article/2010/06/16/nvidia_forceware_25721_driver_performance/1 Outside of a couple of specific scenarios, the overall performance improvement was 1 or 2%, it didn't change anything.

Pretty nice yourself since the article you linked to is 2-3 driver versions old.

 

[Lord_Exodia]

You are freaking wrong. Before I sold my reference 5850 I could get 30% overclock without overvolting and with overvolting I have no doubt i could get 40%+. Look at XS and they have tons of people oc'ing to 1GHz and beyond, many on air.

Comparing oc'd 460s to stock 5850s is rather disingenuous, by the way.

5850 does not overclock further from stock than a GTX 460 1gb period. The most you could get without changing cooling solutions is 5% or less over a 5870 stock performance, from what I've seen and read

Stock clock on a 5850 was 725 I believe on core and to 900-1030ish is the best you can expect.

On the other hand GTX 460 stock clock 675, while people are pushing to 950 even 1000 overvolted with the right mods and cases for cooling.

Now if you can show me several examples of stable 5850 over 1100mhz core I'd agree more.

I think you may be misunderstanding my post. Look at what I'm saying again as I"m not comparing an overclock 460 to a 5850 stock. and saying its a better value. I'm just saying that it can reach 5850 1gb perflormance overclocked sometimes going over 40% faster on some apps with said overclock.

5850 can go higher in clocks but the clock mhz to performance ratio isn't as high and although the overclock can be pushed between 30-35% the actual performance boost in gaming is not as high as a GTX 460 1gb when overclocked.

 

[kllrnohj]

I'll take canned benchmarks any day over "real results". Canned benchmarks are chosen by the developers to represent how the game will play overall. They are always the same with no variables. I much prefer that over someone just running around for 10 minutes. There are way too many variables to fairly compare 2 cards in "real results" for me. For example in metro 2033 there are some areas where just looking two inches higher towards the open sky will drop down 6-7 frames compared to if a player looks dead ahead. Am I suppose to believe the player was able to retrace his steps 100% right down to the camera position through each play through?

Uh, what? Have you EVER seen a canned benchmark? They are about as far from how the game can play as you can possibly get while still using the same engine. Take Crysis, for example. In the canned benchmark, there is zero shooting, zero explosions, almost no enemies or vehicles, no physics being used, and the camera is flying through the air. The real game has nothing in common with the canned benchmark. Most canned benchmarks are they same, they use the card in a completely different way from the game itself.

And the situation hasn't improved since Crysis, Just Cause 2's in game bench is useless and is nothing like the game.

100% reproduction is also not needed. The point is to find out how fast the cards can play games, which real testing does and canned testing doesn't. Even if real testing has a huge margin of error (say, 10% - which if you look at the fraps charts posted, it looks more like 1-2%), that is still infinitely more useful than canned testing, which is 100% irrelevant. Even IF canned benchmarks reflected a similar load to the game, they would STILL be 100% irrelevant the moment driver vendors started doing per-game optimizations. You can bet your ass that both sides optimize the shit out of the canned benchmark first, the game a distant second.

Nice one using an article from May. Its not 20% but it is a very consistent 10% right now. Here is the latest chart from them showing a 9%-13% difference between the 2 cards at every resolution. Strangely enough there is now a 13% gap for your 2560x1600 resolution. http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTS_450_SLI/23.html Magic drivers? Maybe not but a definite improvement.

Improvement in canned benchmarks at least - how about actual gameplay?

 

[hvale2k5]

I have a 470 and it run faster than my friend 5850,

 

[kllrnohj]

5850 does not overclock further from stock than a GTX 460 1gb period. The most you could get without changing cooling solutions is 5% or less over a 5870 stock performance, from what I've seen and read

Stock clock on a 5850 was 725 I believe on core and to 900-1030ish is the best you can expect.

On the other hand GTX 460 stock clock 675, while people are pushing to 950 even 1000 overvolted with the right mods and cases for cooling.

Now if you can show me several examples of stable 5850 over 1100mhz core I'd agree more.

I think you may be misunderstanding my post. Look at what I'm saying again as I"m not comparing an overclock 460 to a 5850 stock. and saying its a better value. I'm just saying that it can reach 5850 1gb perflormance overclocked sometimes going over 40% faster on some apps with said overclock.

5850 can go higher in clocks but the clock mhz to performance ratio isn't as high and although the overclock can be pushed between 30-35% the actual performance boost in gaming is not as high as a GTX 460 1gb when overclocked.

Which doesn't really matter since, as you basically said, the 5850 will still end up faster. The 460 can't OC its way to beating a 5850 if you allow the 5850 to be OC'd as well. 40% faster than stock doesn't matter if it's still slower.

 

[Lord_Exodia]

Which doesn't really matter since, as you basically said, the 5850 will still end up faster. The 460 can't OC its way to beating a 5850 if you allow the 5850 to be OC'd as well. 40% faster than stock doesn't matter if it's still slower.

I was basically disputing what someone said the 5850 is a better overclocker instead of saying the 5850 is going to be faster due to the fact that it too can be overclocked. I never said the 460 is better for the OP

Read the thread and see why someone is posting what they post before you quote them and talk out of your ass.

This is probably the 3rd thread I've seen you quote and respond to someone quoting them and saying that they are saying the exact opposite of what they are trying to say. Do I need to get the Samuel L Jackson Read the Motherfuckin thread before you motherfuckin post and quote reply you

 

[umcpgrad]

people like to twist people's words
what you expect

 

[KipCoo]

I'm just going to wait and see what magic happens when the 6800 series card release. Right now I got a MT Coolermaster 922 case. Found it on Amazon for el cheapos like my burritosm $89 free shipping big baby. Have to say it is a pretty chillsick WCR case. Bout to go to costco here in San Diego and check out what vid cards they got in stock. That's if I still have money left. Just bought a rifle last week.