You do raise a good point, but if the specs are finished by then why not add it in? It will be a nice bonus for the people buying a card for long-term use.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
ATI 58xx series
- Thread starter darkstroke
- Start date
You do raise a good point, but if the specs are finished by then why not add it in? It will be a nice bonus for the people buying a card for long-term use.
Read my mind mate. Will give a advantage to them if nvidia don't have it on theirs and its good for marketing.
Not really as many people use their cards for more than 6 months.
And seeing as its even hard to max out a 4870x2 with the CPU bottlenecking it most of the time a 5870X2 should have enough power for at least 2 years for most & 2 5870x2 should hold my 30" 2560x1600 in good stead for 2 years as i don't like upgrading gfx cards unless i have to & it would be a shame to upgrade through lack of DX when the GPU has the power to hold its own.
Also, if I remember rightly, DX11 is supposed to be coming before Windows 7... Also, I wouldn't be too surprised if devs used DX11, particularly ones who've been working on the 360 because of their prior experience with the hardware tessellation unit (useful because it can artificially increase the amount of triangles that a graphics subsystem can render IIRC).
Last edited:
For exactly that reason, they can't do it properly so they dump it and run off to start again
hope fully they'll make a better job of Dx11
Sounds interesting, but I imagine I'll be skipping to at least the 68xx series and maybe to 78xx. Unless nVidia happen to be on top when I buy, of course. I like the description of how the new x2 cards will be set up though, hopefully it will give worthwhile gains.
Wasn't the same thing said about the 4870x2 pre release, about the cores being stacked? Instead we ended yo with a crossfire sideport which if I remember correctly is dissabled by drivers.
I imagine in future driver releases sideport will be opened up to deliver some of the extra power that the 4870x2 has in reserves?
Either way I hope that this is true but my hopes arn't held to high, I imagine if it can be done, the performance gains from having the increased bandwith between cores would really give Crossfire/SLI the kick it needs to become more mainstream. I say both Crossfire and SLI as you can be sure that one manufacturer will not be heavily outshined by the other, in almost all cases if one does something the other will not be far behind.
I imagine in future driver releases sideport will be opened up to deliver some of the extra power that the 4870x2 has in reserves?
Either way I hope that this is true but my hopes arn't held to high, I imagine if it can be done, the performance gains from having the increased bandwith between cores would really give Crossfire/SLI the kick it needs to become more mainstream. I say both Crossfire and SLI as you can be sure that one manufacturer will not be heavily outshined by the other, in almost all cases if one does something the other will not be far behind.
Man of Honour
- Joined
- 21 Nov 2004
- Posts
- 47,458
Got my 4870X2 and am happy. Will see what the next generation brings...
And yes, the 4870X2 is nowhere near to potential at the moment. Hopefully better drivers will be released in the coming months...
And yes, the 4870X2 is nowhere near to potential at the moment. Hopefully better drivers will be released in the coming months...
- Joined
- 7 Oct 2008
- Posts
- 117
1440x900, but upgrading to 1680x1050 or 1920x1200 at christmas, I'll be getting this card in January
Thanks for all the feedback!
Thanks for all the feedback!
Caporegime
- Joined
- 8 Jul 2003
- Posts
- 30,080
- Location
- In a house
Doesn't really matter does it, we've had Dx10 cards for what, couple of years now or something, and how many Dx10 games have we got, absolute shedload isn't it NOT.
The only real part of that that I really don't believe of it is this:
Since [email protected] effective would deliver about that much bandwidth on a 256-bit bus, which I doubt is out of bounds for higher-end chippery (that and 4870 memory has breached 4GHz).
I mean why would you have 2.3-ishGHz using GDDR5 memory when GDDR3 is cheaper and is easily capable of doing so?
Since [email protected] effective would deliver about that much bandwidth on a 256-bit bus, which I doubt is out of bounds for higher-end chippery (that and 4870 memory has breached 4GHz).
I mean why would you have 2.3-ishGHz using GDDR5 memory when GDDR3 is cheaper and is easily capable of doing so?
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
with a core that small its highly unlikely to have 512mbit interface, not only is it expensive and causes a LOT of extra traces needed on the PCB which again increases cost there as you need multiple more layers to the pcb, like the gtx280 with its eleventy billion layers.
But the pinout for the double sized bus is very large and would add a lot of core size. Its a drop in size, but not massive, its got 25% more shaders but is considerably smaller. Imagine a 55nm with 25% more size, at best would be 320mm, so that down to 205mm is a massive drop if it had the extra bus. Its an unneeded cost that will increase their cost, lower their yields, increase pcb/circuitry/power requirements, etc, etc, etc. Its just not needed considering really, almost a year later there must be some even slightly faster gddr5 available by then.
But the pinout for the double sized bus is very large and would add a lot of core size. Its a drop in size, but not massive, its got 25% more shaders but is considerably smaller. Imagine a 55nm with 25% more size, at best would be 320mm, so that down to 205mm is a massive drop if it had the extra bus. Its an unneeded cost that will increase their cost, lower their yields, increase pcb/circuitry/power requirements, etc, etc, etc. Its just not needed considering really, almost a year later there must be some even slightly faster gddr5 available by then.
Is it? I don't think so.
A 4850 with gddr3 memory at 2Ghz has a bandwidth of 64Gb/sec
A 4850 with gddr5 memory at 3.6Ghz has a bandwidth of 115Gb/sec
For gddr3 memory to even reach 150-160Gb/sec it would need to be running at an "effective" speed of 5Ghz or 2.5Ghz real speed.
I have yet to see any gddr3 memory running higher than 1.2Ghz with volt mods and extreme cooling so not even half of the required bandwidth.
But the figures still don't stack up as you have stated. Although we don't know the speed of the memory, you would have to assume it would be over 1000Mhz which ought to give you in excess of 250Gb/sec with gddr5 and 512 bit bus. It is far more likely then that if the bandwidth is quoted correctly, it will only be a 256 bit bus (much cheaper to make than a 512 bit bus) with the gddr5 memory running at 1.25Ghz real or 5Ghz effective which gives you 160Gb/sec.