ATI HD4800 Review Thread

[Clairvoyant129]

The scaling in that picture is better than perfect. How is that possible?

I understand your point, but look at other games where CF scales really, well near perfection.

But you are right, I'm actually not sure either about how it had better than perfect scaling, I'm just going by what Anand said.

[natty]

I understand your point, but look at other games where CF scales really well, near perfection.

I'm actually not sure about how it had better than perfect scaling, I'm just going by what Anand said.

I'm not debating crossfire scaling. I'm pointing to what looks like an error in that pic. Perfect scaling there would be 111.4 fps for the dual 4870's. And it shows 118.4

[zerazax]

greater than 100% scaling is due to margin of error combination for both single card and dual card tests in the vast majority of cases.

we also tested single card performance on an nvidia system and crossfire performance on an intel system, so the different computers will also add margin of error.

two card solutions generally don't scale at greater than 100% except in extraordinarily odd situations (where rebalancing loads might help with scaling on both individual cards -- but that's odd and rare).

that was posted by Derek Wilson, an editor on Anandtech about the scaling

[RedBull78]

I'm not debating crossfire scaling. I'm pointing to what looks like an error in that pic. Perfect scaling there would be 111.4 fps for the dual 4870's. And it shows 118.4

I think you just want to point out an error in his statement more than what you see in that pic. I think the word perfect should not apply when comes to computuer stuff.

[Unbornchild]

Sorry, we were getting hammered for the last 12 hours. Here is our aftermarket cooler install page.

Wow, that thing occupies what, 5 slots??? Or don't i see it right... do you have a horizontally angled picture how tall that cooler actually is?! Anyway, in a crossfire setup these would be impossible, right?

Just asking out of curiousity.
For myself, i wait for new powerplay results with Cat 8.7... and rivatuner of course.

[BrowncoatGR]

My mobo has one 16x and second 4x PCI-e lanes.. how much performance loss would I encounter if i paired 2x4870's in cf mode?

Probably not a good idea

[junkun13]

should i wait for the 4870X2 or get 2pcs of this now?

[Manicdan]

is it dumb for me to ask if my 2900xt full coverage DD waterblock will work on the 4870? im not scared to mod the block as long as the core and ram are in the same spots.

[RedBull78]

should i wait for the 4870X2 or get 2pcs of this now?

Don't wait if you're not planning to run 4gpu's in your system. As for me, since my mobo has only one true 16x pcie I have no choice but to wait.

[natty]

that was posted by Derek Wilson, an editor on Anandtech about the scaling

why are they putting margin of error into benchmarks? still confused

[RedBull78]

Probably not a good idea

And what about the fact that my mobo doesnt have pcie 2.0?

I can't believe im already behind even tho i upgraded my pc not too long ago...Wow.

[eleeter]

Sorry, we were getting hammered for the last 12 hours. Here is our aftermarket cooler install page.

After reading about 8 reviews, yours was probably the best. Especially mounting an aftermarket HSF, nice info and very helpful.

I think [H] did the worst job. 3 games tested? And they state that is more than enough, I strongly disagree.

[zerazax]

why are they putting margin of error into benchmarks? still confused

Because they probably did more than one run and certain scenarios, the fps in CF was much higher than usual, driving the average of those runs higher and thus appearing to scale > 100%

Either way though, the scaling is apparently excellent in those games

[zerazax]

After reading about 8 reviews, yours was probably the best. Especially mounting an aftermarket HSF, nice info and very helpful.

I think [H] did the worst job. 3 games tested? And they state that is more than enough, I strongly disagree.

I personally thought HardwareCanucks and TechReports were the best overall in terms of technical overview, quality and quantity of games, and settings and so on. Kudos SKYMTL!

[natty]

Because they probably did more than one run and certain scenarios, the fps in CF was much higher than usual, driving the average of those runs higher and thus appearing to scale > 100%

Either way though, the scaling is apparently excellent in those games

Whatever they did it was pretty stupid. People assume they are looking at raw numbers taken from the same exact scenario. And as they admitted, better than 100% scaling only happens in 'extraordinarily odd situations'.

for anandtech

[Eson]

This has to be an error right? Crossfire 4870's scoring over twice as high as a single 4870???

Nothing is identical, but they can be very similar. So what if the second-slave HD4870 were a few percent better then the master card? You would see an increment of over 100%.

[aMp]

Nothing is identical, but they can be very similar. So what if the second-slave HD4870 were a few percent better then the master card? You would see an increment of over 100%.

Not possible unless they were clocked differently or had different timings. This is a methodology issue, not a hardware issue.

[natty]

Nothing is identical, but they can be very similar. So what if the second-slave HD4870 were a few percent better then the master card? You would see an increment of over 100%.

If that was the case then they should put a footnote at the bottom to explain that one card was better than the other. Anyway, I think zerazax already solved this mystery for us.

[eleeter]

Anandtech's testing has been sloppy at times, I think this is another instance of that.

It doesn't take anything away from the 4870, if that is what some people are getting at, that the review is flawed and the 4870 is not all that. Read the other 20 reviews, the vast majority agree that AMD has the best parts in their price category, especially the 4870. It largely invalidates the GTX260, which needs to drop to $300 to level things out.

[zerazax]

That's why I look at all the reviews then make the judgment for myself based on the general consensus. And in the case of the 4870, the general view overwhelmingly is that it matches the GTX260 easily and at $100 cheaper, it is clearly the best buy

[natty]

Yea, I'm not trying to put down 4870, just the pinheads over at Anandtech for artificially inflating results.

[onewingedangel]

well, i looked at the 4870 specs and assumed that it was gonna perform like a 4850.
i was wrong, and i think im gonna buy a visiontek 4870 now.
has anyone overclocked the card yet?

My HD4870 overclocking for the HEXUS review (even though the clock speeds didn't make it into the final review) weren't great 760MHz/4100MHz (yes only 10MHz extra on the core)

Apparantly theres an issue with the HD4870's fan control much like the HD4850, where the fans aren't spinning up like they're meant to so the overclocking will likely improve when a BIOS update is released (as sapphire have told me they've done for the HD4850) or when you use an aftermarket cooler.

[SKYMTL]

Thanks for the compliments guys. Believe it or not, I appreciate the feedback from the community much more than anything a manufacturer can say. Glad to know the review is going to good use.

Wow, that thing occupies what, 5 slots??? Or don't i see it right... do you have a horizontally angled picture how tall that cooler actually is?! Anyway, in a crossfire setup these would be impossible, right?

This depends on quite a few factors such as:

- space between PCI-E slots
- Height of fan
- Orientation of HR-03

You can mount the HR-03 in two ways; one wrapped around the card with the heatpipes "straddling" the card and one like the way I installed it.

It takes up a total of 4 slots.

[SNiiPE_DoGG]

as soon as atitool gets their stuff together well be all set to oc these effectively . I think that once thermal paste is reapplied and the oc's get goon were gonna see a 280 beater in quite a few instances

[865pe]

4870 review

http://techreport.com/articles.x/14990