GeForce GTX 460 1GB SLI vs. Radeon HD 5850 CFX @ [H]

meh you can oc 5850 too why compare soc card to standard 5850 ? nearly every 5850 can go up to 850 or something
 
Has anybody went 5850 to 460 sli?
I ran few tests and seems like my sli is performing slow
ran bfbc2 surround and it feels its running slower than single 5850
and 3dmark06 is giving same result as single 460
16000 with single and sli
15900 with 5850
 
dorky82 said:
Has anybody went 5850 to 460 sli?
I ran few tests and seems like my sli is performing slow
ran bfbc2 surround and it feels its running slower than single 5850
and 3dmark06 is giving same result as single 460
16000 with single and sli
15900 with 5850
Click to expand...

How did you uninstall/install drivers? A full sweep or simple uninstall/install?
And 3DMark06 isn't a good benchmark to compare GPUs, Vantage is better since 3DMark06 ends up being quite tied up with the CPU.
 
eric66 said:
meh you can oc 5850 too why compare soc card to standard 5850 ? nearly every 5850 can go up to 850 or something
Click to expand...

If you read the article, you will discover that they looked at the range of retail 5850's available to find a common factory overclock, then overclocked their CFX'ed 5850's by the same amount.

So, they didn't compare to standard 5850s, and the OCs on both sets were factory OCs and thus officially approved/supported/warrantied.

So the "meh" is back in your court.
 
martinmsj said:
One of the qualities of [H] articles I like is how they bring issues like these (new drivers braking something that worked before) to light. Even going further to to discuss where the results should be once it's fixed. This article doesn't do that, unlike the others I've come to known. Instead, it tells you these are the results and there is nothing abnormal about it. Same goes for AvP if the new drivers broke CFX there too.

Updated my original post a little. Been reading [H] for quite sometime, something is so odd about this article.

Oh well. Had my say.
Click to expand...

We made sure that AMD had notice of the article and the results. AMD verified CFX profiles for the games were included. AMD has informed HardOCP of NOTHING in terms of its drivers being broken.

So when it is all said and done, when you buy the cards, install the latest drivers, and play the game, we show you the performance you will get.

If AMD is going to break its CFX all the time, I am going to leave it in AMD's court to let us know that. HardOCP just tells you guy the real world experience.
 
You know who else realizes that crossfire sucks?

That's right.
Frank Stallone.
 
martinmsj said:
This is the first time I highly questioned a review here.

You didn't mention any issue with 5850 CFX and BC2. So the assumption here is that CFX is working. What I don't understand is how the 5850 CFX is marginally (4-5 FPS) higher than a single 5850. (Mark Warner 460 SLI vs 5850 against Brent Justice 460 SLI vs 5850 CFX.) Saying it's the map map throws my trust out the window for any review now since I can't be sure a review actually represents an experience of a "game."

Even if it's two different Authors there shouldn't be such a disparity between Mark and Brent unless you're telling me now that I have to "not trust one author versus the other."


Even then in the 470 SLI review compared to this one (both by Brent) 460SLI trumps 470 SLI. Heck, in that review the 5850CFX is doing 53.8 with 2XAA trumping the one in your 460 SLI.

Because math is not your strong point?


Sorry Brent, that just doesn't fly. Considering what I've come to believe about the [H] standard, that just throws it all out the window.

*Unless ATI released a profile/driver that killed CFX performance. I wouldn't know I run a single Galaxy 470. Even then, [H] normally points stuff like this out.
Click to expand...

I'm going to have to agree with this post, being told not to compare the same game by different authors for one looks horrible for the [H] review team, and two generally doesn't make any sense. Second, I have plenty of friends who run CrossfireX and play BFBC2 and not one of them have mentioned such a horrible drop in performance thats being implied when you say "AMD broke Crossfire", nore can I find anything online on the subject.

Whats the deal, Kyle? Why don't the numbers add up?
 
Last edited by a moderator:
CHAoS_NiNJA said:
I'm going to have to agree with this post, being told not to compare the same game by different authors for one looks horrible for the [H] review team, and two generally doesn't make any sense. Second, I have plenty of friends who run CrossfireX and play BFBC2 and not one of them have mentioned such a horrible drop in performance thats being implied when you say "AMD broke Crossfire", nore can I find anything online on the subject.

Whats the deal, Kyle? Why don't the numbers add up?
Click to expand...

The statement that 2 different editors can have different real world numbers is correct. If you think about what real world testing really is.

Here's an example; editor 1 runs down toward the base against the wall. A grenade is thrown at him by the enemy and he jumps away, the explosion send bricks flying everywhere, then editor 1 takes out rocket launcher and shoots a missile at the tower to his left and destroys it.

Editor 2 runs toward the base but not near the building, he runs through the grass field, pulls out a machine gun and shoots at his enemies, then takes out a sniper rifle and shoots the guy in the tower, climbs up the tower and picks off the rest of the enemies.

2 different scenarios both could cause different fps from the physx and graphic effect that are used to render the explosions etc...

Hard OCP is correct. You can't compare 2 different people playing, to get 2 people to play identically is almost impossible.
 
Lord_Exodia said:
Hard OCP is correct. You can't compare 2 different people playing, to get 2 people to play identically is almost impossible.
Click to expand...

Thats more then understandable. I, and everyone else looking at the review aren't looking for *exact* numbers, just an explanation for why they are so vastly different. The Crossfire HD5850 numbers on this review are absolutely abysmal, and I can find no where else online that can confirm the same thing. [H]s previous review is an example of how off these current numbers seem.

EDIT: Side note, why does it list Kyle editted my previous post? I see nothing changed in it.
 
CHAoS_NiNJA said:
Thats more then understandable. I, and everyone else looking at the review aren't looking for *exact* numbers, just an explanation for why they are so vastly different. The Crossfire HD5850 numbers on this review are absolutely abysmal, and I can find no where else online that can confirm the same thing. [H]s previous review is an example of how off these current numbers seem.

EDIT: Side note, why does it list Kyle editted my previous post? I see nothing changed in it.
Click to expand...

Kyle added a line into the quote.
 
Last edited:
CHAoS_NiNJA said:
Thats more then understandable. I, and everyone else looking at the review aren't looking for *exact* numbers, just an explanation for why they are so vastly different. The Crossfire HD5850 numbers on this review are absolutely abysmal, and I can find no where else online that can confirm the same thing. [H]s previous review is an example of how off these current numbers seem.

EDIT: Side note, why does it list Kyle editted my previous post? I see nothing changed in it.
Click to expand...

The [H] team explained long ago, that when one editor plays through a scene on the nvidia card he then plays through the scene again on the ati card (no particular order) he tries to play it just like he did before. This way he keeps everything as similar as possible.

What they are saying they cant do is compare how brent plays and how the other guy plays when they do not see each other or play. or they play different scenes or areas, at different times or even days, in different reviews, with different driver sets.

When you see real world numbers on the reviews they were done by one editor with the newest driver sets, on the same day, on the same system (clean install of the drivers and profiles of course) The goal is to measure how an Ati GPU and a Nvidia GPU each handle the situation. They even acknowledge that there is a slight room for error as the enemy AI may not always behave the same and cause slight performance differences.
 
I there, i'am a newb here at HardOCP but anyway.

I was CrossFire 5850 with an OC of 830/1050 (vs 725/1000 normal), that the max i could ever get them.

I recently bought 2 460GTX, (Gigabyte) and i have them running at 822/2050, and if i volt them to 1.087, i get to 880/2000 but this make the fan goes to 80%, make noise and i only gain 1-2FPS.

At 822/2050, not only the cards are cooler at idle but also cooler at Load than my 5850 setup.

I can a much more fluide framerate and way less stuttering in Metro and AvP with SLI

+ i can now use 16xCSAA and all those mode not available on ATI. I can use PhysX in Mafia 2 very well and Batman ;)

So far it's a better experience all along.

I'am only waiting for PowerDVD to patch their player to allow Bitstream HD audio from Bluray disc witch should be coming soon..
 
You must log in or register to reply here.
Back
Top