News Headlines
- Fri, Aug 19
- Final Fantasy XV collector's editions won't include season pass, Square Enix "can't" make more
- Watch Dogs 2 multiplayer will feature seamless co-op and PvP without lobbies or loading screens
- Mario Party: Star Rush gets new box art after making everyone hungry for SpaghettiOs
- Thu, Aug 18
- NVIDIA reveals new GTX 1060 3GB model for $200, taking on the RX 480 4GB
- Capcom to show Monster Hunter Stories and Resident Evil 7 at Tokyo Game Show next month
New Articles
Related Articles
AMD Radeon HD 7950 CrossFire & TriFire Review - PAGE 1
Chris Ledenican - Thursday, February 2nd, 2012 Like ShareThroughout this week we have examined both reference and custom designed Radeon HD 7950 graphics cards, but single GPU performance is only part of the equation. So today we are also going to be looking at both the CrossFire and TriFire performance of the Tahiti Pro GPU.
In order for multiple graphics cards to scale together, the hardware and software (including drivers and games) must all be optimized to support the technology. Since driver support is such a huge piece of the puzzle, AMD has spent a lot of man hours optimizing their drivers and at times will release multiple drivers in a month to improveCrossFire support. Over the last few years, starting with the 5000-series we have begun to see better scaling across multiple GPUs, and we are expecting this trend to continue with the Southern Islands architecture. However, keeping up with the driver support can be a daunting task with new games being released every week.
On the hardware front the Tahiti Pro graphics processor packs quite a punch, as it more or less the 28nm equivalent of the GTX 580. This means it already equipped with a lot of horsepower, so teaming multiple GPUs together should offer performance that is well above graphics cards such as the HD 6990 and GTX 590.

Article Index
|
|

one thing that bugs me about neoseeker reviews is the fact that the actual power consumption of the card itself is not recorded. for nvidia cards, its not a problem, as i can find the info on google in a snap, but AMD cards are a real bitch to find that kind of information on. for a computer builder, it would be more important to know what the card power consumption is, rather than that of the entire system.
I suppose one of the problems with crossfire is that not all games are optimised for it, thus in most cases the extra cost would not be worth it.
Perhaps I'm somewhat expecting too much of AMD, but after the initial issues with the bulldozer CPUs, AMD really could do with this card scoring high, considering that Intel and nVidia perform better in benchmarks. I know a benchmark score might not always translate into good gaming performance, but it seems to be one of the main selling points.
The new architecture seems to be targeted mainly at improving tessellation and gpgpu number crunching, the shader ALU's themselves aren't very different from what they were before.
As for Bulldozer it sold very well despite the reviews. E-tailers are still having a hard time keeping them in stock even though production yields are said to be pretty good.
Hiigaran
TPU does the card only power tests and seem fairly consistent to the neo power consumption if you factor in the other parts.
Does Chris happen to have an Eyefinity setup he can throw these three cards at for benching? Doesn't need to be anything official or testing the other cards but when you hit 60fps+ on nearly all the benches at 2560x1600 it makes it look kinda funny.
The fps on Metro is interesting though, the game has scaled well with tri and quad fire from other reviews I've seen, albeit older generation cards, so I'm wondering if it's bad crossfire support for this gen in the game or it's finally hitting it's single threaded cpu limitation?
any microstuttering, or goofy times with drivers, like for a particular game performance wasnt great right out of the gate, and you needed a hotfix?
also ive never really thought about it, but say you have 3 cards in trifire, each with 1gb of vram. does that effectively give you 3gb of vram, kinda like raid0? does the effective bandwidth also follow in suit? or would you just be stuck with 1gb, as if you only were using one card?