AMD Radeon HD 7900 Reviews

The hosting site haven't been updated with new version of the benchmark.

 

Well, in AMD's defense Q is vastly improved even if HQ still shimmers ( )

 

In the past this site has used out of date performance data for the older cards in the comparison, so beware.

 

The link says it is a hardware error that was carried over architecture over architecture. This confirms the prior assumptions about the nature of the error .

What really annoys me , is how much this problem gets down played by many technicians and reviewers when it was clearly evident in some games , recent of them is Skyrim , IMO something like that should not be ignored .

 

Look what an unlocked BIOS does:

http://www.legitreviews.com/article/1805/14/

 

Rather beastly.

 

The more I read the reviews the more I am liking this card. Still sucks about the comment of AF being < than Nvidias.

 

I am not so sure about the HT4U-videos. Reason: at the end of their image quality section they write the following:
"Damit dürfen wir am Ende des Kapitels Bildqualität attestieren, dass es nun nichts mehr bei der neuen Radeon HD-7900-Serie zu bemängeln gibt."
http://ht4u.net/reviews/2011/amd_radeon_hd_7900_southern_island_test/index11.php
-> At the end of the image quality part, we can thus conclude that there's nothing left to complain about with the new Radeon HD 7900 series.

I've already tried to contact Peter of HT4U, but he seems to be offline - probably recouperating from days with too few hours of sleep.

 

That is the high state clocks. Clock switching can now occur on the UVD state so if the load is low it won't go that high, note that TPU's measurements show playback power draw is lower than 6970. We put clock switching in foe transcode cases where both UVD and the core have heavy processing at the same time.

 

It looks a decent bit faster. It will be interesting to see what NVIDIA does. I'm thinking the graphics card landscape will end up similar to 580 vs. 6970.

 

Nice Hardware.fr test as usual:

http://www.hardware.fr/articles/848-1/amd-radeon-hd-7970-crossfirex-test-28nm-gcn.html

 

Damn good article so far. And it basically proves the F1 2011 CFX/SLI benchmarks from HardwareHeaven are wrong - SLI was disabled.

 

Is it just me or does performance tank more heavily than in case of other cards when MSAA is applied?


EDIT:
I just looked at MSRP here in Poland, the price is exorbitant to say the least... I thought perhaps I'd go for it, but no, thank you for now.

 

Well, big chip on a brand new process with low initial yields and no immediate competition just won't sell for pennies.

 

Yeah, I know. It costs twice as much as the 5870 I last bought though. Pretty steep.

Oh well, gonna wait for Kepler, perhaps there'll be more options then. Unless I get bored waiting .

 

It would be more interesting at least for me to see how the announcement of this product pushes down the prices for the rest of the lineups from Nvidia and ATI of course.

 

A couple things from the quote from AMD in the Anandtech review (http://www.anandtech.com/show/5261/amd-radeon-hd-7970-review/12)

From the last part, it's pretty evident SSAA isn't coming to DX10+ in AMD drivers anytime soon.

But more interestingly, in the first part (the bolded part), it looks like we may have some new driver level AA options soon.

At 549, the cad is entirely way too expensive for me. So while I was ready to buy it on release due to ZeroCore and increased perf over my 5870, the price is just too steep for me. And with Nvidia unlikely to launch a better performing part in the next 4-6 months, I have a feeling the price won't change much for a while.

Regards,
SB

 

5870 itself held for quite long time its MSRP, even after Fermi launch. So if the 28nm supply is going to be poor, especially for large die yields, the prices could hold up well within 2H'12.

 

One thig I forgot to add. It appears that AMD are greatly increasing their budget for ISV relations. I'd imagine that isn't going to be cheap so maybe the greatly increased price is going to go to pay for better developer relations. If so, then that isn't necessarily a bad thing...but it still puts the card out of my budget range.

Perhaps if it had a greater performance jump or even at least supported PLP for Eyefinity gaming I could try to squeeze it in there. But basically 550 USD is a high price to pay. I'd never give Nvidia that much for one of their GPUs, and it's no different for AMD.

Regards,
SB