Nice kit, but I won't be spending that much on a graphics card again any time soon..
The one criticism I would levy is that it's not a 2GB card - it's two crossfire 1GB cards. Unless an application with 2GB textures can hold it all on the card, it's not 2GB..
PK
I must say one thing for the 5970 - they at least managed to keep the launch date under wraps pretty well, for a change
I have to admit I wasn't expecting it to be out nearly as soon, mainly because of the supply issues than anything else I guess though.Originally Posted by scaryjim
I must say one thing for the 5970 - they at least managed to keep the launch date under wraps pretty well, for a change
That'll probably hold true for at least a month or two anyway I suppose :/
I love it when a big release like this just happens with no warning and you're left scrambling to try and see what you've missed
OK I will probably sound like an idiot asking this but it's a question that I always wondered:
If ATi (and nVidia) can pair two of their graphics cards together on the same PCB, and be able to deal with the subsequent problems of power draw and heat as well as the obvious engineering problems, then why can't they release a single monolithic card with the same specs? (i.e., double the specs of the 5870 in a monolithic card rather than this CrossFire business).
Is it because of the yield issues of getting a die that big? That's my very vague understanding anyway so can someone please help
I wondered if they'd hold it back until something speculative came out about a Fermi release date, just to try to "rain" on NVidia's parade. But unless ATI know something the rest of us don't, I guess not. (unless, of course, they *are* planning a 5990 at much higher clock speeds, and that's the one they're saving up...)
Isn't that basically what NVidia keep doing?
Also, if the roadmap I've seen knocking around is anything to go by, the top-end Fermi will have 512 "cuda cores" - more than twice a GTX285... perhaps that's why NVidia are having so many problems getting production-grade Fermi cores...?
From a CUDA course I took, it turns out that a lot of the chip is interconnects between cores. Add more cores = more interconnects, lower yields. The nVidia architecture isn't as scaleable as the ATi architecture, but it's easier to program in.
٩(̾●̮̮̃̾•̃̾)۶
Curious statement, do you really think CUDA is any easier to code for than OpenCL or DirectCompute? I had a browse through the CUDA headers a few weeks ago, and what I saw was not pretty.
Dabbled a bit in CUDA for work, and Stream in my spare time, and CUDA seemed easier for me. It suits my programming style. Not tried OpenCL yet, waiting 'til it comes out of beta properly.
٩(̾●̮̮̃̾•̃̾)۶
Fair enough. Just wondering.
Can you imagine being the guy in charge of tracking out a card like that? Thinking to yourself "even if I use about 50 PCB layers and spend 6 years routing it, it's still going to be LOLBIG".
PHP Code:$s = new signature();
$s->sarcasm()->intellect()->font('Courier New')->display();
I'm sure they've got fairly robust auto-tracer software, for designing them. The number of traces will still be insane though, so manual tweaks is going to be an issue.
Well at over £500+ I'm going to wait until Nvidia's cards make a show cause these prices are just to rich.
I dislike Nvidias practices at this minute but ATI seem to be rolling over the market and taking full advantage of the fact. Kind of sad in my opinion.
too expensive, the games out today dont deserve such amount of money to be spend
I'm gonna wait for NVidia's Fermi response. Not that I'm a NV fanboy, but I just wanna make sure that if I'm spending £500 I want to be able to compare both products!!
There are currently 1 users browsing this thread. (0 members and 1 guests)
Posting Permissions
LinkBack URL
About LinkBacks
Reply With Quote
