ATI HD4800 Review Thread

[SKYMTL]

Would be nice if some other site tried a 4870CF on p45 vs x38 comparison. Or better yet 4870x2 CFX when they come out to see if there are any bottlenecks

This in itself is very hard to do. Let me explain.

Due to the different natures of the P45 and boards with X38 / X48 chipsets it is almost impossible to have a completely even playing field. Heck even when testing boards WITH THE SAME CHIPSET there will be differences in gaming performance. I will draw your attention to a few articles:

Techreport's X48 Showdown (See ET: EQ benches)

Bit Tech's Multi GPU benches with the Gigabyte P45 DQ6

OC3D's P45 Showdown

Not only are there performance differences between X48 and X38 (basically the same chipset) when equipped with similar hardware but there are even VERY large differences between P45 boards when it comes to performance. We aren't talking about 1-2%, more like 10-20% difference in some cases.

So, the question begs to be asked: in a Crossfire vs Crossfire review between the X38 / X48 and the P45, how do we know that the performance (or lack thereof) of the motherboard itself and NOT the PCI-E bandwidth is being brought to the forefront? The simple answer is that there is really no way to know unless there are MASSIVE differences in performance.

Writing a review of Crossfire performance between two motherboards with different chipsets, getting results that are within 5-10% difference between the two products and then saying that there is a bottleneck is a stretch at best. That is why I look at some preformance results and take them with a grain of salt.

Will a P45 bottleneck a pair of HD4870 cards? Heck no.

[biohead]

This in itself is very hard to do. Let me explain.

Due to the different natures of the P45 and boards with X38 / X48 chipsets it is almost impossible to have a completely even playing field. Heck even when testing boards WITH THE SAME CHIPSET there will be differences in gaming performance. I will draw your attention to a few articles:

Techreport's X48 Showdown (See ET: EQ benches)

Bit Tech's Multi GPU benches with the Gigabyte P45 DQ6

OC3D's P45 Showdown

Not only are there performance differences between X48 and X38 (basically the same chipset) when equipped with similar hardware but there are even VERY large differences between P45 boards when it comes to performance. We aren't talking about 1-2%, more like 10-20% difference in some cases.

So, the question begs to be asked: in a Crossfire vs Crossfire review between the X38 / X48 and the P45, how do we know that the performance (or lack thereof) of the motherboard itself and NOT the PCI-E bandwidth is being brought to the forefront? The simple answer is that there is really no way to know unless there are MASSIVE differences in performance.

Writing a review of Crossfire performance between two motherboards with different chipsets, getting results that are within 5-10% difference between the two products and then saying that there is a bottleneck is a stretch at best. That is why I look at some preformance results and take them with a grain of salt.

Will a P45 bottleneck a pair of HD4870 cards? Heck no.

Dam good to know, sir.

[Caveman787]

Why didn't they save all the 4850 reviews and bring the to this sticky?

[L7R]

Will a P45 bottleneck a pair of HD4870 cards? Heck no.

Thanks for clearing this. I Just read that topic in Hardware canucks forum too.

[SKYMTL]

Thanks for clearing this. I Just read that topic in Hardware canucks forum too.

I don't know if I cleared it up...seems like I just raised more questions in my own head.

[Xcel]

How big a difference is there if the 4870 is run at pice1 16x instead of 2.0?

[SKYMTL]

How big a difference is there if the 4870 is run at pice1 16x instead of 2.0?

This is in the same vein as what we have been discussing and truth be told, there is really no good way of finding out due to the differences in the mobos.

[Xcel]

This is in the same vein as what we have been discussing and truth be told, there is really no good way of finding out due to the differences in the mobos.

Sorry I couldn't bother reading the entire thread. So in other words pretty minimal then if it's within the error spread.

*continues reading the rest of the thread*

[mascaras]

hi guys


just installed a powercolor 4870






Idle temp with FAN Fix = 45% (ambient temp = 23º)

[SKYMTL]

Mascaras, you do realize that your idle temp will be high due in part to your use of Aero, right? Just ran a quick test on my stock HD4870 and found turning off Aero shaved about 5°C off idle temps.

[SNiiPE_DoGG]

Mascaras, you do realize that your idle temp will be high due in part to your use of Aero, right? Just ran a quick test on my stock HD4870 and found turning off Aero shaved about 5°C off idle temps.

as long as its not up at 80 im happy ^^ 840 core and i idle at 47C, fortunately my airconditioner running next to me is louder than the card will ever be I think applying mx-2 to the core is a good idea for everyone

(ambient is 74F)

[Shadout]

Is there any way to force a X38/X48 motherboard to only use x8 instead of x16 PCI-E. I guess there isnt :/

[SKYMTL]

I haven't found a way.

[mascaras]

E8400@4Ghz
HD4870 Default
Catalyst 8.6 hotfix
Crysis
windows vista 64bit
1680x1050
DX9
High settings

Average = 45 Fps

[SKYMTL]

I would suggest you retest with the Assault_Harbor timedemo included with that benchmark suite and run it two times. The first run should usually be discounted since it takes into account the frames per second on the Load screen which will of course lower your score.

[MpG]

PCI-E Scaling Review (THG)

I'm curious, does anyone know if this is a valid way of deactivating lanes?

[ThatGuy16]

mascaras has me drooling over one of these things. Only one problem though, I'm broke.

[emoners]

Is there any way to force a X38/X48 motherboard to only use x8 instead of x16 PCI-E. I guess there isnt :/

why would you wanna do that?

[ripken204]

mascaras has me drooling over one of these things. Only one problem though, I'm broke.

ya same here
but i can afford the card, im just waiting for a waterblock!

[Shadout]

why would you wanna do that?

I assume it would make it a hell of a lot easier to test x8 vs x16 bandwidth on the new GPUs.
Would remove the variable of using both P45 and X38 for testing.

[FischOderAal]

Is there any way to force a X38/X48 motherboard to only use x8 instead of x16 PCI-E. I guess there isnt :/

Dunno, maybe you can put tape on some of the connectors of your gfx or something. At least for RAM this was possible. Some did this to revive their DIMMs after the SPD was accidently deleted by some gigabyte mainboard.

[mascaras]

Diamond HD 4870 XOC Premium Black Edition: First Picture

As we reported earlier here Diamond Multimedia is preparing a special unlocked Black Edition Radeon HD 4870 card. We just received a first picture:





http://www.techpowerup.com/65100/Dia...t_Picture.html

[Direwolf]

If the diamond version would cost only $50 more It would be nice. I would guess they raised the voltage a good deal on the gpu/memory for more overclocking goodness.

For the lazy people not wanting a VMod

[zerazax]

Throw 1GB on that sucker and you'll have quite the competition to the 280

[Caveman787]

I predict this card may take over the gtx280 and the 4870x2 will take over the gtx280 ultra or match it.

I also predict since we've seen amazing performance from ATI's card's on the higher end that all the lower end ones are gonna kick the crap outta nvidia's versions